Introduction to a Catalogue of the Tibetan Manuscripts from Tun-Huang in the India Office Library

Louis de la Vallée Poussin

Even if the texts contained in our manuscripts had been complete and furnished with tolerably clear colophons, the task of arranging them in a reasonable and practical order would have been difficult. It is well known that the Tibetan lo-tsa-vas were not always able to decide whether a praṇidhāna is a Vinaya, Sūtra, or Tantra: the same text is sometimes to be found in two parts of the Bkaḥ-ḥgyur. Even the distinction between Bkaḥ-ḥgyur and Bstan-ḥgyur is not in every case certain. To give a few instances — the Aṣṭamaṇḍalaka is both Mdo and Rgyud; the Prajñā-pāramitā-hṛdaya is both Śer-phyin and Rgyud; the Bhadracarī is both Ḥdul-ba and Rgyud, but its commentary is Mdo-ḥgrel.

To the essential difficulties of arrangement of Buddhist matter are to be added the difficulties due to the state of a large part of the collection, and, last but not least, to my own narrow limitations. It is, for instance, beyond my power to ascertain whether an unidentified 'tāntrik' fragment is Mahāyāna-sūtra or Tantra; whether a 'śāstra' fragment is properly called Śāstra (a translation from an Indian treatise) or Siddhānta (Śāstra of Tibetan origin). We possess such Siddhānta (the works of Dharmasiddha for instance); and in some cases affinities with the Mañjughoṣahāra-vajra's Siddhānta and quotations of Sūtras known to be rather unfamiliar to Sanskrit authors give some evidence in favour of Tibetan authorship. Therefore, while aiming at a scientific arrangement, I have chiefly endeavoured to make researches as easy as possible. From many points of view this catalogue is a tentative work. The following order has been adopted:

  1. 1-47 Vinaya texts (Hīnayāna)
  2. 48-204 Sūtras and commentaries on Sūtras, with identified Sanskrit titles
  3. 205-225 Sūtras translated from Chinese or with Tibetan title, but without ascertained Sanskrit title
  4. 226-305 Sūtras and commentaries, unidentified fragments, arranged according to (1) the topics and (2) the characters (Upāya-kauśalya and other kauśalya, 226; karman, 234; kāya, 235; tathatā, tathāgata, 238; prajñā (cosmology), 239; pudgala-dṛṣṭi, 242; bodhicitta, 243; bodhisattva practices, 244).
  5. 306-455 Tāntrik works with identified Sanskrit titles, including properly so-called Tantras and Dhāraṇīs (Rgyud) with matter that is properly Rgyud-ḥgrel: Sādhana, Vidhi, Tāntrik Śāstra, Pūjā ritual including Stotra, Pariṇāmanā, Praṇidhāna, excepting the Bhadracarī
  6. 456-71 Tantras and Tāntrik matter, identified texts but without ascertained Sanskrit titles
  7. 472-587 Tāntrik works, unidentified fragments arranged according to the topics
  8. 588-652 Śāstras with identified titles
  9. 653-85 Śāstras, unidentified, arranged in the following order: Abhidharma topics, Mahāyāna topics, Mādhyamika philosophy, Yogācāra topics (Bodhisattva-caryā), Vijñānavādin philosophy
  10. 686-765 Tibetan works: with ascertained titles; unidentified, on the scriptures and the schools, on dogma, history, divination, medicine

The unidentified works are arranged, on the whole, according to the matter; the identified works in alphabetical order of the Sanskrit titles, manuscripts containing more than one work being placed according to the first. The topics of the unidentified works are indexed in the Sanskrit index.

Among the Vinaya texts may be mentioned: (a) Extracts from the Vinaya-vastu (1, 2, 3), vastu being translated dṅos-po, not gźi as in the Bkaḥ-ḥgyur, and extracts dealing with Pravrajyā-vastu (3, 4), which may conveniently be placed here. (b) A number of copies of the Prātimokṣa-sūtra, differing from the Sanskrit text from Turkestan and the Kumārajīva translation. Two translations (5, 8, 9 and 6, 7, 10) differing one from another and both differing from the Bkaḥ-ḥgyur, the differences being very slight. The most conspicuous is that the summaries are introduced by mdor-na or by sdom-ni. (c) Fragments dealing with Vinaya topics (24, 25), parts of treatises, summaries of the Sutta-vibhaṅga or of commentaries on the Prātimokṣa or kinds of Nidāna-pudgala-saṃgraha: their exact nature has not been ascertained. 23, 26 contain fragments of ecclesiastical history (sources, one Chinese, are indicated). (d) A commentary on Guṇaprabha's Vinaya-sūtra (34). (e) Two works not traced in the Bstan-ḥgyur, Vinaya-saṃgraha-stotra-bhāsya (42) and Vidhi of the reception of the vows of a lay devotee (43). The texts dealing with the Mahāyāna or Bodhisattva-vinaya have been catalogued among the Sūtras or Śāstras. The most important will be found at nos. 54, 244, 245, 250, 262, 626, 632, 633, 634, 635, 636, 675, 676.

Among the Sūtras special attention may be claimed for the Acintya-prabhāsa (49), where the title does not agree with the Bkaḥ-ḥgyur translation acintya; Daśasaṃjñāna (89, 311, 4), where there are really eleven notions; the Bkaḥ-ḥgyur names the Sūtra 'Saṃjñānaikādaśa'; different translations of the Prajñā-pāramitā-sañcaya-gāthā (110, 111, 112, 113, 114, 115, 116); the Māra-damana (155), not traced in the Bkaḥ-ḥgyur, which may be a chapter of a big Sūtra; the Ratna-ketu-dhāraṇī (156, 157, 158, 159, 160) where the title does not always agree with the Bkaḥ-ḥgyur translation of ratna.

While the collection does not include more than forty-five different identified Sūtras, the tāntrik manuscripts are much more numerous, and the task of arranging them is very difficult. It will be seen that a large number of the titles have not been traced in the Tibetan canon, but it would be imprudent to jump to the conclusion that they are new. Specially interesting are a commentary on the Thabs-kyi-źags-pa-pad-moḥi-phreṅ (321); a collection of ritual texts (376); the Uṣṇīṣa literature (322, 323, 324, 325, 351, 352, 353, 354, 355, 356, 357, 358, 359); and the Mahābala literature (389, 390, 391, 392, 393).

Among the Śāstras may be noted translations from the Chinese (588, 589, 595, 619, 620, 625); recensions of the Bodhicaryāvatāra attributed to Akṣayamati (628, 629, 630); works not traced in the Bstan-ḥgyur (602, 609, 616, 626, 640, 641, 644, 645, 647, 648) some of which may be Tibetan works (and there are probably Sanskrit works catalogued as Tibetan).

The colophons often give the titles of the work and the names of the translators; often too, when the work is a Śāstra, the name of the author. The colophons of the Aparimitāyur-jñāna-sūtra (310) and of the Prajñā-pāramitā (104, 105, 107, 109, 117) also name the scribes. For convenience other colophons — not many — which contain the names of scribes or any interesting notice are listed here: 120 ('Written for the merit of paṇḍita Devaghosa'), 260, 311, 4 ('...the undertaking of Prajñāśrī is completed'), 321, 334, 372, 380, 1 ('Translated in the bird-year in the twin-palace (?) by the translator Dharmatāśīla'). It will be seen from 104, 105, 109, that the formula ...źus seems to be equivalent to the formula ... źu-chen-bgyis.

A double principle has been adopted in the choice of extracts from the texts: to provide the reader with passages susceptible of being identified or of literary interest (comparison with the Tibetan canon or Sanskrit texts), and secondly with passages which have not been understood. Nor has it been forgotten that the orthography of the manuscripts is likely to be interesting. Passages written in red are distinguished by asterisks at the beginning and end. The irregularities mentioned below, under Orthographical Features, have accordingly been reproduced without comment.


All paper, except Ch.73.III, small fragments of birch-bark each containing only one or two characters: 502 is a block-print with Chinese pagination; 765 is written on modern Russian paper. The make-up of the manuscripts has been described as pothī, concertina, roll, scroll, and book-form.

  1. Pothī. The folios vary greatly in size — the largest being those of the Prajñā-pāramitā texts: they may have one or two holes, with black, yellow or red circles: they may be unruled, or with black, yellow or red lines, with vertical lines left and right or without. The paper may be thin or thick, or of two leaves stuck together, with or without water-marks. The texts in pothīs are Sūtra, Vinaya or Tantra and are as a rule carefully written. The mūla in commentaries is often written in red (indicated by *....* in the extracts quoted).
  2. Concertina. Long sheets of rather thick paper, or double paper, folded, written continuously on recto and verso, leaves unnumbered, of moderate size. As a rule they contain Tāntrik works, inaccurately written.
  3. Roll. Made up from sheets of different length, each of which is divided by ruled lines into two or three pages, unnumbered; written on one side, parallel to the length of the whole roll. With a few exceptions, the texts are Aparimitāyur or Prajñā-pāramitā, accurately written.
  4. Scroll. Made up of large leaves of paper, written continuously, without division into pages, parallel to the width: the larger scrolls are mounted on thin Chinese paper (often cut from several Chinese texts), inaccurately written.
  5. Book-form. In 420 the leaves are joined at the top, elsewhere in the European way.

The rolls of the Aparimitāyur-jñāna-sūtra (310) deserve special notice. The paper is quite different from the paper of the other manuscripts in the collection. The majority consists of three sheets, each with two pages, stuck together to make a roll of six pages: some have seven pages, an additional half sheet having been added. All the copies examined agree with the two edited by Professor Sten Konow (Hoernle, Manuscripts Remains, Hoernle_1916: pp. 289 et seq.) in omitting 8 to 31 of the Hoernle edition. Some copies have been corrected in red ink, and may record the scribes, the same names occurring more than once.


The leaves of the pothīs are usually numbered, verso, on the left — seldom on the right — margin; sometimes recto (e.g. 592, where there is no consistency).

  1. By far the most usual mode of pagination is the Tibetan numbering gcig, gnyis... (indicated by 'numerals' in the descriptions). 62 has bcig bcig, gnyis gnyis...
  2. A prefixed letter sometimes indicates the volume, as in the case of the big Prajñā-pāramitā manuscripts, or the work (565); frequently (619 and elsewhere) the hundred. It is sometimes doubtful whether the letter refers to the volume or the hundred (e.g. 588).
  3. Very often the scribe uses letter-numerals, ka 1 — a 30. 552 has ka, 'kha, ga...
  4. Or he uses letter-numerals and numerals ka gcig, kha gnyis...
  5. When there are more than 30 folios, it is necessary to use a prefixed or sub- or superscribed letter to distinguish the series of thirties.
  6. In 82 the series runs: ka gcig....a sum bcu, ang so gcig.
  7. The letter-numeral series may omit a; for instance in 615 kha ha is followed by nga ka.
  8. In 67 two features are worthy of notice: the letter ma is subscribed to the letter-numeral and the series begins with a (a+ma The manuscripts affords no clue to the meaning of the subscribed ma. Other manuscripts confront us with obscure riddles.
  9. 641 has both letter-numerals and letters: da kh+nga, na ng+ma, pa kh+ma (11, 12, 13?)
  10. 113 (a single folio) is numbered k+ka go bnyi. We might admit three series of 1-31, 32-62, 63-93, but that is not very likely.
  11. 206 has folio numbered k+ka. The earlier folios are numbered with the ordinary letter-numerals and three is some ground for believing that the value of the double ka is 31 (but no certitude, far from it).
  12. 382 provides a double pagination, 126-37:
  13. 48 begins with ordinary letter-numerals (folios 7-13, 19-20): then part of a series with superscribed ra (or nga); then part of a series with superscribed sa, and finally one folio with a subscribed ra (gra). A previous numbering is cancelled in the superscribed sa series.
  14. 645 consists of two folios, numbered ka, ki.
  15. 651 has Tibetan letter-numerals and Chinese numbers (referring to the work).
  16. Khotanī numbers in Tibetan transcription are used in 336, 338, 340, 423, 424, 425, 553, sometimes with Tibetan letter-numerals. 336 (five folios) has gcig, 'dwa, gsum, bzhi, co ra.


'a (see Laufer, Bird Divination, Laufer_1914a: pp. 52 et seq.; Barnett, Ancient Khotan, Stein_1907: i. 549). It appears as 'a and 'a [with hook] both forms occurring in the same manuscript. Sometimes, but rarely, omitted (nam mkh, Laufer_1914a: p. 55). Its use below the line to mark a long syllable in Sanskrit transcriptions is seldom accurate. Frequently used finally, even after i, e, and o (na' = na : pa' = pa : dge' = dge : so' = so, etc.), but inconsistently in the same manuscript (e.g. 206, 252, 284, 465, 668, 669, 679). It occurs much less frequently in the middle of a word, except in 'da's and gna's which are fairly common (e.g. 206, 273, 465, 648). sems dpa's occurs in 206, ston pa'r byed in 456. 252 also shows a prefixed 'a in bsgrub 'pa'i ye shes. Other uses worth notice are the transcriptions 'ge'u ta ma (49) : a sang 'ga' (613); 'di rga a ga ma (690) and omissions in pi, bi = pa'i, ba'i (93, 670) and poe = pa'o (56).

To sum up, it is doubtful whether our manuscripts support Laufer's opinion that 'writing was then in its initial stage; and the rule as to when the letter a was a necessity, and when it could be dispensed with, was not yet clearly developed'.

As concerns two peculiarities of the unclassical spelling, the da-drag and the ya-btags (Laufer_1914a: pp. 57 and 96; Barnett, Ancient Khotan, Stein_1907: i. 559, and Barnett_1903: p. 110), nowhere are they carefully applied. The da-drag is in full swing in a large number of manuscripts, but consistently practised in few; it is absolutely wanting in some (e.g. 300). According to Barnett it was already beginning to be dropped in actual speech, and according to Laufer there was no hard-and-fast rule for its application in writing and it was no proof of antiquity itself. The ya-btags is much more common and, it may be said, the rule.

The classical distinction of gi gyi : gis gyis is ignored, and there is confusion in the use of kaga : cacha : tathada : paphaba : tsatsha, the classical and irregular uses appearing side by side in the same manuscript at times (e.g. 206, 257, 284, 446, 465, 590, 648, 653, 659, 668). The assimilation gag gis = gang gis is fairly common: 273 has gag de'i = gang de'i. Also fairly common is rga for dka: for instance in 670 rgon and dkon are used indiscriminately. The classical la sogs (ādi) is written la stsogs (e.g. 600, 604, 655, 668); las stsogs (e.g. 604, 616, 648); la bstsogs (681); las bstsogs (689.2); las bsogs and la brtsogs (670).


The writing is described as dbu-can and dbu-med, although real dbu-med with the ligatures given by Csoma is not common. The dbu-can is of the well-known type, more or less cursive: 'square' dbu-can occurs in a few manuscripts.

As is well known, i is written i and I [reversed]. Many scholars believe that the different forms correspond to phonetic variants, either in quantity or 'timbre'. The impression is that they are used indiscriminately, for practical purposes, an explanation that seems sufficient. The different ways of writing Om, Hūm, Svāhā are interesting (e.g. 97, 317, 516, 548, 558). The forms of the letters are the usual ones, except for 'a (noted above), and for ra superscribed, which is often very like nga (e.g. 48, 221). Nga and na are sometimes subscribed in yang, dang, and yin, usually when words occur at the end of line, but not always.

Abbreviations like ngagi = ngag gi are not rare (see also suppression of 'a in bi above).

Anusvāra is rare except bījas, although thaMs= thams is common enough.

In punctuation the classical rule — no dot before the stroke except after nga — is usually ignored. A number of manuscripts have two dots : before the stroke and a single dot after every syllable; in others one or two dots are used indiscriminately. A dot before the genitive suffix 'i is frequent, even after u and o.

To sum up, I am certainly not tempted to overestimate the importance and antiquity of the present collection. It was natural that the discoveries in Central Asia should have been greeted with enthusiasm, and their linguistic and archaeological interest is in fact enormous; but there is no reason why everything from Central Asia should be very interesting. I lack the competence necessary for understanding and criticizing a large part of the collection, namely the pure Tibetan works; the catalogue required the endeavours of a Tibetan scholar, whereas I claim only some knowledge of philosophical and Sūtrik Tibetan.

I do not believe that there is as yet a Tibetan palaeography. As far as I can judge, the use of the da-drag and ya-btags, the use and abuse of 'a, and that sort of thing, are not proofs of a high antiquity. I have no title to form an opinion on the nature or the date of the paper used, but the presence in the collection of a sheet of nineteenth-century European paper shows that modern documents may have found their way into the celebrated cave library.

The great majority of works which I have been able to identify have been traced in the Bkaḥ-ḥgyur Bstan-ḥgyur. They are attributed to the same translators, when the translators are given, in the manuscripts. Two of the colophons state that the translation is in brda-gsar, new style. But there are many divergences from the canon, and in a few cases different translations of the same texts, which are very interesting and constitute a direct proof of the antiquity of the manuscripts. For instance, 96 translates adhiṣṭhāna byin kyi rlabs, 97 has gnas. Folio 34 of 112 gives also an indication of the progress realized by the new translators.

It is a duty to state that I am largely indebted to the Catalogue of the Bstan-ḥgyur of Cordier Cordier_1909_1915, and also to the advice and the numerous direct and side lights with which Dr. F. W. Thomas has provided me.