Vous êtes sur la page 1sur 14

Luc Kwanten

The Lexicography of the Hsi Hsia (Tangut) Language

In: Cahiers de linguistique - Asie orientale, vol. 11 n2, 1982. pp. 55-67.

Rsum Depuis la dcouverte d'une importante bibliothque Hsi Hsia il y a bientt soixantequnze ans, l'tude de cette langue morte n'a gure progress. Parmi les problmes qui demeurent, le plus important est l'absence d'une reconstruction satisfaisante du systme phontique de la langue. Il n'existe qu'un document bilingue, le Fan-Han ho-shih chang-chung-chu (1190). Les reconstructions phonologiques qu'on en a tires, fondes sur des usages chinois rcents, ne sont pas valables. Il faudrait utiliser nos connaissances sur les dialectes chinois du nord-ouest au Xme sicle. En second lieu, il est ncessaire de compiler un dictionnaire fiable - ce que ne sont, pas les trois ouvrages existants, deux russes et un japonais. Pour mener bien une tude lexicographique correcte, il faut d'une part tudier le seul texte bilingue dont nous disposions, d'autre part faire une analyse comparative (avec les originaux chinois) de traductions en Hsi Hsia de textes chinois tels que, par exemple, le Lun-yu ou le Meng-tzu. Jusqu' prsent, on ne dispose que d'une seule tude de ce type, sur L'Art de la Guerre de Sun-tzu. La comparaison du glossaire contenu dans ce travail avec les dictionnaires existants confirme l'inadquation de ceux-ci. Cette inadquation rend problmatique la lecture des textes linguistiques en Hsi Hsia sur lesquels on a fond les reconstructions existantes. L'auteur donne quelques exemples tirs du Chang-chun-chu .

Citer ce document / Cite this document : Kwanten Luc. The Lexicography of the Hsi Hsia (Tangut) Language. In: Cahiers de linguistique - Asie orientale, vol. 11 n2, 1982. pp. 55-67. doi : 10.3406/clao.1982.1115 http://www.persee.fr/web/revues/home/prescript/article/clao_0153-3320_1982_num_11_2_1115







hsi hsia

(tangut) language ( )

Mahmud be name the al-Kasjfari's is established, the in the On referred language within that in The name usually explained cultural tradition plural of years This yet the the state own still frequently (1) al-Kasgari Tang%ud. people developed year 982 unique ofMongols the , Turkic of as in and accepted. state, of Diwan History on and, today. in explanation group re occurrence and part been the puzzle first the northwesternMongol to of its 300 the first around Secret its state most HsiMongol main a ethnic nearly the inscriptions the the the Luvat of occurs being the for century tradition not of remains Tang-hsiang. Although lasted border cannot Sung Formed (960-1272), Tang\ud has that name eight Chinese name state of a centuryat-Turk. the the etymology Orkhon are as In eleventh Hsia fact, basis ; a the


to the Tang^ud as "a tribe of Turks, they live near China"(2).

(*) Revised version of a paper presented at the 15th International Conference on Sino-Tibetan Languages and Linguistics, Beijing, People's Republic of China, August 1982. I wish to thank Prof. K.J. Weintraub, Dean of the Humanities, The University of Chi cago, for assistance in attending this conference. (1) E. Haenisch, Die Geheime Geschichte der Mongolen, Leipzig, 1948, para. 265 ff. (2) Mahmud al-Kastfari, Compendium of Turkic Dialects (Diwan luyat at-) , Vol. 2, ed. and transi., R. Dankoff, Sources of Oriental Languages and Literatures 7, (forthcoming). Talat Tekin, A Grammar of Orkhon Turkic, Bloomington, IN, 1968, p. 277 ff. Cahiers de Linguistique Asie Orientale Vol. XIN2Dcembre 1982, pp. 55-57



From its inception, the Hsi Hsia state maintained very close ties with the Ch'i-tan state to the north of the Sung. It was under Ch'i-tan influence that the second ruler of Hsi Hsia, Li Te-ming (1004-1032), apparently the most scholarly of the Hsi Hsia rulers, began to develop a complex script, apparently using both the Ch'i-tan and the Chinese script as his models. During the reign of his son Li Yuan-hao (1032-1048), the developing script was standardized in a manner reminiscent of Li Ssu's ac tions on the Chinese script during the Ch'in dynasty (B.C. 22 1 207) (3). Unlike many of the so-called invented scripts such as, for example, Ch'i-tan and 'Phags-pa, the Tangut script was in active use until at least the second decade of the fourteenth century. A voluminous literature developed which included Buddhist, his torical, legal and economic texts. Many Chinese, and possibly also Uighur texts were translated and written down in a script that can be easily be termed the most complex ever developed in East or Central Asia. Of all the destruction wrought upon civilization by the Mon gol onslaught, the annihilation of the Hsi Hsia state and its cul ture was the most thorough. A limited knowledge of the language remained until at least the beginning of the fourteenth century. In 1307, the Yuan dynasty, for reasons that still remain to be in vestigated, published an edition in some 3,620 chapters of the Buddhist Tripitaka in the Tangut script (4). By the late 133O's, however, it had become, for all practical purposes a dead language. There is no indication in the contemporary sources that the few Yuan officials of Tangut origin were capable of reading the Tan gut script. By the time of the Ming dynasty (1368-1644), the lan guage was not only a dead one but also a forget ten one. Ming and Ch'ing scholars remembered the existence of a Hsi Hsia state and compiled several works on its history ; the language and the script, however, had fallen into oblivion. It was not until the latter part of the nineteenth century that the language was rediscovered. In 1870, A. Wylie undertook a study of the unknown script on the multilingual inscription of the Chu-yung-kuan(5) , an inscription dated 1345. Although Wylie (3) L. Kwanten, "Tangut Miscellanea. I. On the Inventor of the Tangut Script", Journal of the American Oriental Society, vol. 97, 1977, pp. 333-335. (4) K.T. Wu, "Chinese Printing under Four Alien Dynasties, Harvard Journal of Asiatic Studies, vol. 13, 1950, pp. 447-523. (5) The best study of the Chu-yung-kuan is provided by Murata Jiro, Kyoyokan, 2 vols., Kyoto, 1955-1958. The facsimile of the inscription has been provided by R. Bonaparte, Documents de l'poque mongole du XlIIe et XIV sicles, Paris, 1885 and in which Tangut has been identified with a question mark. A. Wylie, "On an Ancient Buddhist Inscription at Keu-yung-kwan" , Journal of the Royal Asiatic Society, 1870, pp. 14-44.



was unable to identify the script, he suggested that, most likely, it was that of the Jurchen Chin dynasty (1115-1234). This identi fication remained accepted until 1882. At about that time, the French scholar G. Deve'ria undertook a study of the Yen-t'ai stele in Honan which he published in the Revue de l 'Extrme-Orient (6) . This inscription was clearly, and beyond doubt, identified on the stele itself as being in the Jurchen script. The unknown script of the Chu-yung-kuan in no way ressembled it. Deveria's suggestion that the latter was the script of the Hsi Hsia or the Tangut state led to an animated controversy between him and the eminent French sinologist E. Chavannes. G. Deveria persisted and in 1895 he wrote a communication, not published until 1901, en titled "L'criture du royaume Si-hia ou Tangout"(7) in which he proved the existence of a specific and unique script, quite dif ferent in appearance from the Jurcen script. He had come to this conclusion by examining the bilingual inscription on the Kan-ying tower in Liang-chou, once an important Tangut city(8) . As a direct consequence of this study, several other articles identifying material in the Tangut script were published. Nearly all of them concerned bilingual Chinese-Tangut numismatic and epigraphical material but, at no time, did anyone succeed in transla ting Tangut text. The first such effort was published in 1904 the by G. Morisse who, while with the French legation in Peking, had acquired a Tangut translation of the Saddharmapundarka(9) . Using a Chinese version as his guide, Morisse was able to provide the pronunciation and the meaning of a small number of Tangut charac ters well as a sketchy grammatical outline. as Most likely, Tangut studies would never have developed beyond this preliminary stage simply because very little material written in Tangut was known to exist. The beginning of the twentieth cen tury, however, was an era of massive European exploration of China's borderlands and of Central Asia. In 1908, the Russian ex plorer P.K. Kozlov explored the Qaraqoto area in present day Kansu province and discovered in the ruins of the city a hidden li brary of material in the Tangut script. This material was crated (6) G. Devria, "Examen de la stle de Yen-tai", Revue de l'Ex trme-Orient, vol. 1, 1882, pp. 173-186. (7) G. Devria, "L'criture du royaume Si-hia ou Tangout", Mmoires, Acadmie des Inscriptions et Bettes Lettres, Institut de France, vol. IX, pt. 1, 1901, pp. 147-175, (8) G. Devria, "La stle Si-hia de Leang-tchou", Journal Asiatique, 1898, pp. 53-74. (9) G. Morisse, "Contributions prliminaires l'tude de l'criture et de la langue Si-Hia", Mmoires, Acadmie des Inscriptions et Belles Lettres, Institut de France, 1e srie, pt. 2, vol, 11, 1904, pp. 313-362.



and shipped to Leningrad, then St. Petersburg, where it is still housed in the library of the Institut Vostokovedeniy AN SSSR. Kozlov conducted two excavations in Qaraqoto, respectively in 1908 and 1909, and in his report to the Imperial Geographical Society, the sponsor of the expedition, he stated : "We have made a find of great scientific value and discovered a true trea sure. The expedition has found thousands of complete books and a quantity of scrolls, individual volumes and pages as well as hundreds of Buddha representations in painting and sculptures"(10) . The importance of the Kzlov discovery, which raised the known inventory of Tangut materials from less than a dozen items to several hundreds, was confirmed by the Russian sinologist A. Ivanov in a communication to the Imperial Academy of Sciences in St Pe tersburg on December 1, 1909 (November 18) and entitled "Zur Kenntnis der Si-Hia Sprache"(1 1) .With this communication, serious stu dy of the Tangut language had finally begun. The topic of Ivanov1 s communication was an extremely impor tant document, namely the Fan-Han ho-shih chang-chung-chu , a Tangut-Chinese and Chinese-Tangut glossary in which each entry was accompanied by a Tangut as well as a Chinese phonetic symbol. Thus, for the first time, both the meaning and the pronunciation of Tan gut characters could be derived from a native text, a text dated 1198. Until then, all attempts at phonetic reconstruction had to be based on Sanskrit and were tentative at best. Although Ivanov1 s communication marked the first serious step forward in the study of the Tangut language, the Chang-chung-ohu itself was not published until some twenty years later when Lo Fuch'eng issued a handwritten copy of it in Tientsin(12) . Privately published, this copy was not readily available to other scholars. A consequence of this delay in its publication was that any stu dies had to be based on the slightly more that 300 words that Ivanov had included in his communication, without providing the Tangut original. Unfortunately, Ivanov made several serious mistakes in his reading of the Chang-ehung-chu. First of all, most likely as a consequence of the undeveloped state of Chinese historical phone tics, Ivanov had the Chinese phonetic glosses transcribed in the Peking dialect of the nineteenth century when these symbols re(10) K. Kozlov, Izvestiy a Imperatorskoi Geograf . Obsc. , vol.45, 1909, p. 429. The paintings, nearly all frescoes, and the sculp tures are on exhibit at the Hermitage Museum, Leningrad and have been studied by S.F. 01'denburg, "Materiali po buddiiskoi ikonografi Kharakhoto", Materiali po Etnografii Rossii, St. Pe tersburg, 1914, pp. 79-155. (11) A. Ivanov, "Zur Kenntnis der Si-Hia Sprache", Izvestiya Impera torskoi Akademii Nauk, ser. VI, vol. Ill, nos 12-18, 1909, pp. 1221-1233. (12) Lo Fu-ch'eng, Fan-Han ho-shih chang-chung-chu, Tientsin, 1928.



fleeted twelfth century Chinese as spoken in the northwest. Se7 cond, Ivanov read the characters from left to right rather than the correct right to left. Ivanov' s error was not discovered un til nearly eighteen years later (13). The most important study in Tangut linguistics, still con sidered as such today, was published by Berthold Laufer(14). In this study, on the basis of the Chinese phonetic transcriptions, he compared Tangut with Lolo and Moso and concluded that these languages were closely related. The main consequence of Laufer1 s study was the classification of the Tangut language in the SinoTibetan group where, rightly or wrongly, it has remained ever since. As I have pointed out during the 14th Sino-Tibetan conference and in my recent study of the phonology of the language as reflec ted the Chang-ohung-chu, there are some serious questions as to in the validity of what may be called the Laufer hypothesis ( 15) . In deed, Laufer used the wordlist published by Ivanov and although he criticized Ivanov1 s use of the Peking dialect for this phonetic reconstructions, he was unable to notice Ivanov1 s erroneous read ings. Laufer proceeded to transcribe these same glosses read in a similar manner in the contemporary "mandarin" dialect which was not any closer to the correct reading than that by Ivanov. Fur thermore, he started from the premise that Tangut had to be a Sino-Tibetan language because of Chinese statements on the ori gins of the Tang-hsiang. In proceeding in this manner, Laufer ignored not only the history of the Tangut people but also the historical evolution of the Chinese language. In addition to the errors made by Ivanov, Laufer added an important one of his own. He tacitly dismissed the pronunciation difference between twelfth and twentieth century Chinese and thus compared a language of the twelfth century with Lolo and Moso using only nineteenth century phonological data(16). (13) E. von Zach, "Uber einen storenden Fehler in den bisherigen Si-hia Studien", Orientalistisohe Literaturzeitung , vol. 30, 1927, pp. 4-5. (14) B. Laufer, "The Si-hia language. A Study in Indo-Chinese Phi lology", T'oung PaOy vol. 17, 1916, pp. 1-126. (15) L. Kwanten, The Timely Pearl. A Twelfth Century Bilingual Chinese - Tangut Glossary , vol. I, The Chinese Glosses , vol. II, The Tangut Glosses and Dictionary , Uralic and Al taic Series, vols 141-142, Indiana University, Bloomington, IN (in press) ; L. Kwanter, "The Phonological Hypothesis of the Hsi Hsia Language", T'oung Pao, (Forthcoming). (16) The Lolo and Moso data used by Laufer was recorded rather haphazardly and not by trained linguists so that this data, and hence the comparison, is of an even more dubious validity.



The Laufer hypothesis became the theoretical basis for two different reconstructions of the language on this hypothesis, a Russian and a Japanese one(17). These reconstructions have been made without a re-examination of the Laufer hypothesis, without examining the data that can be derived from a proper analysis of the Chang-chung-chu and without considering the fact that, alrea dy 1927, E. von Zach raised serious doubts about the validity in of the research hypothesis. My own study has but amplified the doubts that should have existed with regards to the Laufer hypo thesis and, indirectly, about the validity of the current affi liation of the Hsi Hsia or Tangut language. My examination of the Laufer study, however, concerned not merely the validiy of the phonological reconstructions proposed by him, and subsequently by Nishida Tatsuo and M.V. Sofronov, but also the validity of his etymological comparisons. By means of these comparisons, Laufer linked Hsi Hsia to Lolo and Moso, languages spoken in an area geographically remote from the lo cation of the Hsi Hsia state. It is evident that if the phonological reconstruction is in error, etymological comparisons based on that reconstruction will also be inaccurate, especially when, in addition, a tempo ral and spatial continuity is assumed. The Laufer etymologies, however, suffer not only from phonological errors but also from some very basic lexicographical mistakes that render his compa rative study, and thus the affiliation with Lolo and Moso, all but useless. Although Laufer had correctly realized the value of the Chang-ehung-ehUy his inability to examine the text first hand prevented him from realizing the fundamental mistakes made by Ivanov. Both Ivanov and Laufer, and at present Nishida, assumed that Hsi Hsia characters were structured in a manner identical with that of Chinese characters. They assumed that, for example, the character for "man" always had the same form regardless of its semantic or grammatical function. The examination of the Chang-ahung-ahu reveals that this is not the case. There are at least four Hsi Hsia characters to express the notion of man, each of which is translated by the Chinese jen A. , but the form of the Hsi Hsia character varies with its grammatical function(18). The re-examination of the phonological hypothesis on which Sofronov and Nishida based their reconstructions, quickly revealed (17) M.V. Sofronov, Grammatika Tangutskogo Yazyka, 2 vols,, Moscow, 1968 ; Nishida Tatsuo, Seikago no Kenkyu, 2 vols., Tokyo, 1964-1968. (18) Nishida Tatsuo, The Structure of the Hsi Hsia (Tangut) Cha racters, James A. Matisoff, Transi., Kyoto, 1979. Additional examples are provided in L. Kwanten, "The Phonological Hypo thesis of the Hsi Hsia Language", op, cit. .



the need for a similar re-examination of the known lexicography of the language. Before proceeding with this topic, however, it is important to keep in mind that, although archeology has re vealed only one bilingual glossary, the Chang-chung-chu y we have at our disposal a treasure trove of materials translated from Chinese and of which the original is known. Hence, a Roset ta stone of sorts is available and proper lexicographical studies, albeit mute ones, can be undertaken. As I lave pointed out be fore, this work has to be accomplished before any comparative etymological studies can even be considered. The first dictionary of the Hsi Hsia language was compiled by N.A. Nevsky in the 1930's but was not published until 1960(19). It is a facsimile of his work notes and we will return to it la ter. Prior to this publication, three word lists had been pu blished. The first one by G. Morisse in 1904, i.e. before the Qaraqoto discoveries ; the second one by Ivanov in 1909 and a third one by Nevsky in 1926. These lists, however, have a very limited lexicographical value as their principal concern was the phonetic reconstruction of the language. G. Morisse examined the Hsi Hsia version of the Saddhavmapundavka sutra and compared it with its Chinese version(20) . In so doing, he was able to establish a small dictionary of some 100 characters with Sanskrit phonemes, the majority of them per sonal names. At that time, however, very few Hsi Hsia texts were available and no serious attempts at translation were undertaken. Two bilingual epigraphical texts, the only two other texts known to exist, were studied on the basis of their Chinese text with the assumption that the content of the Hsi Hsia text to be identical(21). When A. Ivanov examined the Hsi Hsia material that had been send to St. Petersbourg, he did discover the Chang-chung-ahu and immediately understood its significance. A few months later, in November 1909, he presented a paper to the Imperial Academy, the sponsor of the Kozlov expedition, in which he described the do cument, provided a list of 300 words, their translation and their pronunciation in the modern Peking dialect. Thus, Ivanov' s paper could be considered the first "dictionary" of Hsi Hsia, if it had not been riddled with some very basic mistakes. The Chang-chung-chu is a work that anticipates the much bet ter known Huai i-yu except that, in addition to having been com piled by non-Chinese, it provides not only Chinese phonetic glosses (19) N.A. Nevsky, "Tangutskaya Filologiya. Issledovaniya i slovar" , 2 vols., Moscow, 1960. (20) See note 9. (21) G. Devria, "La Stle Si-hia de Leang-tcheou", Journal Asia tique, 1889, pp. 53-74. In addition to the Liang-chou stele, there was the multilingual inscription of the Chang-chung-chu.

62 Lua KWANTEN for the Hsi Hsia characters but also Hsi Hsia phonetic glosses for the Chinese characters. The mistake made by Ivanov, and in this he was unwittingly followed by Laufer, was to assume a ho rizontal correspondence between the Chinese and the Hsi Hsia characters. In fact, Ivanov had postulated that the grammatical structure of Hsi Hsia was identical to that of Chinese. A mere cursory examination of the document reveals that this is far from the truth and that many of Ivanov' s "translations" are to tally inaccurate. Nevertheless, it was on this list, with minor modifications in the phonology, that Laufer based his comparative etymologies. The foremost lexicographer of the language was N.A. Nevsky, a Russian scholar active in the 1920's and 193O's(22). Better than anyone else, Nevsky realized that before a proper phonetic reconstruction of the language could be attempted, scholars needed a dictionary, albeit it a partially mute one. As was stated above, his work - never completed - was not published until 1960. The on ly partially lexicographical work published by Nevsky was a study of Hsi Hsia characters with Tibetan transcriptions. Although this work contains some 334 characters, it provides but few definitions and, hence, could not be used for translation work. Its purpose was primarily phonological (23) . Among the Hsi Hsia material recovered in Qaraqoto were a num ber of works that appeared to be concerned with the language as such. Although these works are best known by their reconstructed Chinese titles, it should be pointed out that, unlike the Changchung-chUy they contain not a single Chinese character and are completely mute. These works are known as the T'ung-yin> the Wen-hai, the Wen-hai pao-yun, the Wen-hai tsa-lei and the Wusheng ch'ieh-yim. It is commonly assumed that the Wen-hai is a phonological dictionary based on a Chinese model and that it covers 97 rhymes of the level tone, whereas the Wen-hai pao- covers 86 rhymes of the rising tone. The Wen-hai tsa-lei is viewed as a dictionary of miscellaneous rhymes. The T'ung-yin, the principal work on which the current phonetic reconstructions are based, is consi dered to be a dictionray of homonyms, and the Ch'ieh-yun appears to have been a sound table of Chinese for Tangut usage. This lat ter work, which would be extremely valuable to have, appears to have been lost again. (22) His biography (cf. L.L. Gromskovskaya and E.I. Kychanov, Nikolai Alexandrovich Nevskii, Moscow, 1978) does not state much about the end of his career or the circumstances of his death. It is mentioned in Solzhenitsyn, The Gulag Archipelago. (23) N.A. Nevsky, A Brief Manual of the Si-Hia Characters with Tibetan Transcriptions y Osaka, 1926. The original documents on which this study was based appear to have been lost again.



The correct lexicographical study of the languages has to . proceed from a comparative study of the Hsi Hsia translations of Chinese originals and other languages, if they can be found. Up to now, only one such study has been made available, namely Sun Tzu's Art of f/ar(24). The comparison of the partial glos sary in this work with the so-called dictionaries reveals the glaring inadequacies of the latter. It casts a severe doubt on our ability to read technical, linguistic works such as the Wenhai and the T'ung-yin. An attempt to study these works through translations based on erroneous lexicographical data will result in the fact that any phonological reconstruction, or any theore tical model for such an endeavor, will be severely deficient. Since the publication of Nevsky's slovar , two other "dic tionaries" have been published. The first was compiled by Nishida Tatsuo on the basis of his phonological study of the T 'ung-y in (25) . It contains some 4,000 characters with definitions, in Japanese, and a proposed phonetic reconstruction. It is about 1700 charac ters shorter than the character list compiled by M.V. Sofronov on the basis of the same document(26) . Sofronov's work contains no definition, only a proposed phonetic reconstruction. Whereas Nishida goes in great detail into the theory upon which he bases his phono logical reconstruction, he does not specify how he ascertained the meaning of the characters without recourse to another text trans lated from Chinese. The second dictionary, published in 1968, was a translation into Russian of the Wen-hai by a team of Soviet scholars(27) . This work, known as the More Pis 'men and unlike the work of Nishida, contains a facsimile of the material translated. The facsimile contains 3064 characters, whereas the Russian work contains 5145 characters. In most instances, translations are given as well as some phonological indications. Like Nishida1 s work, the Russian scholars do not state upon what basis they have translated this document. Hence in both cases we are asked to accept their definitions, or translations, at face value. Elsewhere, I have extensively examined the hypothesis with regards to the nature of the Wen-hai and the T'ung-yin and there is no need to repeat what I have stated there (28) . Suffice it to (24) K.B. Keping, Sun' Tzi'v Tangutskom Perevode, Moscow, 1979, especially pp. 383-453. (25) Nishida Tatsuo, (26) M.V. op. cit., vol. 2, pp. 309-505. Sofronov, op. ait., vol. 2, pp. 276-403.

(27) K.B. Keping, V.S. Kolokolov, E.I. Kychanov and A. P. Terent1 ev-katanskogo, More Pis 'men. Faksimile Tangutskikh Ksilografov, 2 vols., Moscow, 1969. (28) See, L. Kwanten, "The Phonological Hypothesis of the Hsia language" 9 op. ait. Hsi-

64 Luc KWANTEN say that the extant interpretations of these two works is invalid and that work on them has to be started anew. Nevertheless, it is these two works that have given us our two principal "dictionaries". It is, therefore, imperative that these "dictionaries" be re-exa mined from a lexicographical viewpoint and, as was the case in the phonological examination, the Chang-ohung-chu serves a our primary source(29). The Chang-ohung-chu contains some 1200 characters of which 75 % can be translated with absolute certainty, i.e. they occur more than once in the text. I selected 875 of these characters for a detailed comparison with the work of Nishida and with the More Pis 'men. 63 characters were listed neither by Nishida nor by the More Pis 'men y the latter failed to list an additional 135 characters and the former 28, thus leaving 650 characters for exa mination. The More Pis 'men did not provide 284 characters with a definition while Nishida failed to provide a definition for only 3 characters, leaving a total of 363 characters that could proper ly examined. be Of the 363 characters translated both by Nishida and the More Pis'men, 133 characters or 37 % are in complete contradiction with each other, leaving but 230 characters, or less than 27 % of the characters initially selected. When these 230 characters are compared with the definitions determined on the basis of the Changohung-chu , a full 50 % are inaccurate. Extrapolating, it would leave Nishida' s dictionary with approximately 520 accurate entries, and the More Pis 'men with about 668 accurate entries, figures that render any translation of a text nearly completely impossible. From what precedes, it is obvious that, as was the case with the phonological reconstruction, the work that has been done un til now is in need of serious revision and emendations. It is al so obvious that with an accuracy rate of less than 15 %, together with erroneous phonetic reconstructions, we are not in a position, at this time, to undertake scholarly justified comparative etymo logical studies. The only dictionary that proves to be usable, is the one pro vided by Nevsky's work no tes (30). His dictionary contains about 4,000 characters and they are all based on either the Chang-chungchu, or comparative studies of Hsi Hsia translations of the Chinese Confucian classics or selected Chinese buddhist sutras. In each case, Nevsky quotes the example and hence the usage of the charac ter becomes immediately clear. It is, however, a posthumous publi cation of an unfinished and unrevised work, hence a substantial number of entries have no definition at all, or definitions that cannot fully be attested. In addition to this, the work is entirely (29) See, L. Kwanten, The Timely Pearly vol. II, The Tangut Glosses and Dictionary , op. oit. (30) N.A. Nevskii, Tangutshaya Filologiya, vol. vol. 2, pp. 7-666. I, pp. 173-601 ;


manuscript and Nevsky's cursive Russian is very difficult to read. Still, about 2,000 characters can be used without too much dif ficulties. It is, nevertheless, not possible to use this work for comparative etymological purposes as Nevsky treated the structure of Hsi Hsia as identical with that of Chinese and, hence, did not provide the semantic variants of individual characters. The correct lexicography of the Hsi Hsia language presents numerous problems to the scholar, and herafter I will give but a few examples based on the Chang-chung-ahu. Example 1 Chinese : aheng lE Hsi Hsia Jfaft , 1 occurrence, Nishida 204-108 H ( ), More Pis'men ; term for the first of spring. Hsi Hsia text : ' More HsiaH45)f : , Virtuous, just, Nishida 051-053 Ih. , Hsi Pis'men 2 occurrences, behaving justly, correctly, truthHsi^ia text : a)i* $ b)# ft. 5 ) 11

More Hsia vfc which occurrence, Nishida 007-1 1a JPfc ^ , merging, Hsi admixture , 1 is truly, really uniform, blending, an Pis'men 1255 : not maintained. Hsi Hsia text s J^. 7J Probable translations : Hsi Hsiafljj^fc Hsi Hsia^l^ Hsi Hsia apt Example 2 The word for grass, Chinese ^ ts'ao is represented by two characters in Hsi Hsia, namely : $. and 3>}^| respectively glossed as /)) ^f and shih/eig "jL . (31) The second entry is 10th century northwest Chinese as recons tructed by Lo Ch'ang-p^i, T'ang Wu-tai hsi- farig-yiin, Shanhai, 1932. Until we have a reconstruction of 12th century north western Chinese, we have to postulate a 10th century pronuncia tion 12th century glosses. for Chinese loan for TJ, equivalent to Chinese ML. l^. first (in time) J$T ^-

in official titles



The first one is Nishida 144-031 *? , the second one is Nishida 185-064 w ; in the More Pis'men the numbers are 0464 "general term for every possible grass, term for various sorts of grassy plants", and 0573 term for grass. Occurrences : 1 each in the following expression :

f k

determine the of the Hsi Hsia lexeme for grass, not yet comparative On the basis not possible. studies are basic available evidence, it is hence possible to Exemple 3 The same applies to determining the lexeme for the number one. There are three Hsi Hsia characters, namely ^j , , >' occur once in the following expressions :




work again, with the current state of our knowledge, comparative One is not possible. Example 4 The Hsi Hsia character \& is listed by Nishida 004-081 as a personal name. It is not listen in the Move Pis'men. The de finition provided by the Chang-chung-chu is W .to see. Example 5 The Hsi Hsia character ^fj, is listed as More Pis'men 1920 "a Chinese word for wooden lattice/grating". Also used for crushing legs (as an instrument of torture) . Nishida 005-082 defines it as an olive tree. The Chang-chung-chu provides as a definition G) a mortar, a bowl commonly used for preparing rice. This lat ter meaning is in perfect accord with the section in which it is found, namely eating utensils. From these few examples, it can be noted that, as was the case with Hsi Hsia phonology, Hsi Hsia lexicography is in great need of revisions and corrections. This work should proceed from the word list of the Chang-chung-chu provided in my study of the document, through a comparative study of materials translated

THE LEXICOGRAPHY OF THE HSI HSIA (TANGUT) LANGUAGE from Chinese such as, for example, the Lun-yu(32) . It is only when we have an adequate dictionary, and a correctly reconstruc ted phonology that comparative etymologies can be undertaken. Until then, care has to be exerted in the use of Hsi Hsia ma terials for comparative Sino-Tibetan studies.


Lue KWANTEN The University of Chicago RESUME Depuis la dcouverte d'une importante bibliothque Hsi Hsia il y a bientt soixante quinze ans, l'tude de cette langue morte n'a gure progress Parmi les problmes qui demeurent, le plus important est l'absence d'une reconstruction satisfaisante du systme phontique de la langue. Il n'existe qu'un document bilingue, chang-chung-chufl-ZPOj . le Fan-Han ho-shih

Les reconstructions phonologiques qu'on en a tires, fondes sur des usages chinois rcents, ne sont pas valables. Il faudrait utiliser nos connaissances sur les dialectes chinoi du nord-ouest au Xme sicle. En second lieu, il est ncessaire de compiler un dictionnaire fiable - ce que ne sont, pas les trois ouvrages existants, deux russes et un japonais. Pour mener bien une tude lexicographique correcte, il faut d'une part tudier le seul texte bilingue dont nous disposions, d'autre part faire une analyse comparative (avec les originaux chinois) de traductions en Hsi Hsia de textes chi nois tels que, par exemple, le Lun-yu ou le Meng-tzu. Jusqu' prsent, on ne dispose que d'une seule tude de ce type, sur L'Art de la Guerre de Sun-tzu. La comparaison du glos saire contenu dans ce travail avec les dictionnaires existants confirme l'inadquation de ceux-ci. Cette inadquation rend pro blmatique la lecture des textes linguistiques en Hsi Hsia sur lesquels on a fond les reconstructions existantes. L'auteur donne quelques exemples tirs du Chang-chun-chu .

(32) I have completed a translation and study of the Hsi Hsia translation of the Confucian Analects and hope to soon pre sent it in a published form.