Bian Xiang

8/13/2019 Bian Xiang

1/83

Harvard-Yenching Institute

What is Bianxiang? -On The Relationship Between Dunhuang Art and Dunhuang LiteratureAuthor(s): Wu HungSource: Harvard Journal of Asiatic Studies, Vol. 52, No. 1 (Jun., 1992), pp. 111-192Published by: Harvard-Yenching InstituteStable URL: http://www.jstor.org/stable/2719330

Accessed: 22/05/2010 08:29

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at

http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless

you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you

may use content in the JSTOR archive only for your personal, non-commercial use.

Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at

http://www.jstor.org/action/showPublisher?publisherCode=hyi.

Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed

page of such transmission.

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of

content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms

of scholarship. For more information about JSTOR, please contact [email protected].

Harvard-Yenching Instituteis collaborating with JSTOR to digitize, preserve and extend access toHarvard

Journal of Asiatic Studies.

http://www.jstor.org
http://www.jstor.org/stable/2719330?origin=JSTOR-pdfhttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/action/showPublisher?publisherCode=hyihttp://www.jstor.org/action/showPublisher?publisherCode=hyihttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/stable/2719330?origin=JSTOR-pdf


2/83


3/83

112 WU HUNG'transformed' into 'popular storytelling.' "2 Sun Kaidi *,MMlikewise suggested that the character bian had the same meaning inboth bianwen and bianxiang but that it should be interpreted as''miraculous transformations" recorded in religious texts: "Whensuch figures and events are described in writings people called thembianwen, or bian for short. When these are depicted in pictures peo-ple called them bianxiang, or again, bian for short. Their meaningsare the same."' Both opinions have been followed and refined byscholars. The relationship between bianwen and bianxianghas alsobeen considered in terms of bianwenperformance. For example, SunKaidi noted that the "Wang Zhaojun IEE- Bianwen" from Dun-huang ended with the phrase: "The end of the 'standing-pu' vza ofthe first juan *; from here the following part of the text begins."Since the term pu was a measure word for art works includingpaintings, he contended, "this makes it clear that a popular lecturehad pictorial images as aids."5The etymological and functional approaches were synthesized byFu Yunzi 44r# in an important proposal: "Originally bianwensup-plemented (i.e., explained) bianxiang pictures. . . . Bianwen andbianxiangare identical in implication but different in their methodsof presentation: one uses words while the other uses pictorial im-ages. Pictorial spatial representations are bianxiang; and oral tem-poral presentations are bianwen."6 To support his argument, Fulisted subjects that appeared in both bianwen iterature and bianxiangmurals, such as the "Hell bian" *I-AN, the "Bian on the subjuga-tion of demons" RNW, the "Mahamaudgalyayana bian" HAW,

2 Zheng Zhenduo, Zhongguouwenxuehi rPEJJS;4t, 2 vols. (Changsha: Shangwu yin-shuguan, 1938), 1:190.3 Sun Kaidi, "Du bianwen erze" *-II3ii MIJ,riginallywritten in 1936, a revisedversionis in Sun Kaidi, Cangzhouji IJIt- (Beijing: Zhonghuashuju, 1965), rpt. in Dunhuang ianwenlunwenlu#MWUWPr'Q, ed. Zhou Shaoliang J0V and Bai Huawen W4Lt, 2 vols.(Shanghai: Shanghai guji chubanshe, 1982) [hereafterLWL], 1:241.4 An introductionto different interpretationsof biancan be found in Mair, T'angTransfor-mationTexts,pp. 36-72. Mair himself insists that bian n both terms, bianwen nd bianxiang,means "supernaturaltransformation."5 Sun Kaidi, "Jindai xiqu yuanchu Song kuileixi yingxi kao" fIN M l;qH5AM

WC FurenXuezhi ZfE t 11.1/2 (1942): 37.6 Fu Yunzi, "Sujiang xinkao" fXT~ , originallyappearedin Xinsichao1.2 (1946): 39-41; LWL 1:154.


4/83

WHAT IS BIANXIANG? 113and the "Vimalakirti bian" SAN. He then deduced that theseliterary and artistic works "originally formed pairs." Going onestep further, he proposed that, in addition to these fixed wall paint-ings, illustrations in a portable scroll format could have also beenused in the telling and singing of Buddhist bianwen; so suggestedthe Dunhuang manuscript "Bianwen on Mahamaudgalyayana's res-cue of his mother from hell" (S.2614), which was inscribed withthe words: "with pictures, one scroll." Moreover, he argued thatpaintings were also employed in telling non-Buddhist bianwenstories, as suggested byJi Shilao's &1=i05poem, which describes asinging girl from Sichuan performing the Zhaojun bian (WRkit"t) with "a picture scroll."Written half a century ago, Fu Yunzi's discussion contains all themajor arguments of some recent scholars who, in attempting todefine bianwen precisely, consider the interdependence of bianwenand bianxiangto be one of bianwen's essential features. ReiteratingFu's view, Cheng Yizhong ;-Vt claimed that "bianwen are ex-planatory texts of bianxiangpictures."7 His proposal was supportedby Bai Huawen. After investigating some twenty surviving Dun-huang manuscripts specifically entitled bianwen, Bai asserted thata bianwen must possess two primary characteristics: a prosimetricstructure and a bianxiang picture as a visual aid in its performance.The first feature is visible in the text itself. The second is indicatedby an extant Dunhuang scroll painting illustrating the bianwen on"Subjugation of demons"; by the terms tu (X, picture) or pu Nfound in some titles or texts of bianwen; and by phrases regular-ly appearing at the junction between spoken and sung parts in abianwen text that often "direct the audience to 'look' at the sametime as they listen."8 Bai's strategy and conclusion have been ac-cepted and further developed by Victor Mair,9 in whose enriched

7Cheng Yizhong, "Guanyu bianwen de jidian tansuo" W J I, ,*, originallyappearedin Wenxueyichanengkan MA:IR V 10 (1962): 80-101; LWL 1:373.8 Bai Huawen, "Shenmo shi bianwen" ffM jtgI, LWL 1:429-45. English translationby Victor Mair, "What Is 'Pien-wen'?"HJAS 44. 2 (1984): 503. Unless noted, the followingreferencesto Bai's work refer to the English version.9 Among a number of definitions of bianwen, he one Mair "hold(s) ultimately to be theonly truly workableone" "results in a corpus of less than 20 extant bianwenmanuscripts."See Mair, T'ang Transformationexts,p. 14.


5/83

114 WU HUNGformulation of bianwen's characteristics we again find "an implicitor explicit relationship to illustrations."'"Then, what are bianxiangaccording to these authors? After defin-ing picture scrolls used in bianwenperformances as bianxiang,Bai fur-ther addresses the following question: given that bianwen texts fromDunhuang far outnumber such scrolls, "How then, may they besaid to go together?""1 In response, he points out that the ThousandBuddha Caves at Dunhuang are covered with bianxiangmurals andmany banners from the caves are also painted with narrative scenes.Again following Fu Yunzi, he concludes: "Not only could bianwenbe used in combination with painted scrolls for general, secularperformances, they could also be used in combination with wall-paintings, painted banners, and so forth for performances in Bud-dhist temples, in stone grotto chapels, or in certain specially desig-nated situations." 12

Unsatisfied with Bai's reasoning, Mair has made an admirableeffort to collect all the ancient literary evidence for bianxiang.3 Onthe basis of this research, Mair both agreed and disagreed withBai's contention: on the one hand, he realized that this data failedto yield any specific proof for the performing function of bianxiangmurals;"4 on the other, he insisted on the "narrative nature" ofbianxiang pictures, which permitted him to translate the Chineseterm bianxiangas "transformation tableaux":Probably most importantly in terms of its true identity, pien-hsiang [bianxiang] mustbe recognized as a kind of narrative art. From literary descriptions and first-handobservation of painting and sculpture, we may induce that, while various devices

10 Other basic featuresof a bianwenn Mair's definitioninclude: "a unique verse-introduc-tory(or pre-verse)formula, an episodic narrative progression, homogeneity of language,and prosimetric structure." See ibid., p. 15.Bai Huawen, p. 505.12 Ibid., p. 507.13 Victor Mair, "Records of transformation ableaux(pien-hsiang), TP 72.3 (1986): 3-43.14 Without ruling out a possible relationshipbetween bianxiangnd bianwen erformance,

Mair cautions his readers, very correctly in my opinion, not to take such a relationshipforgranted: "We have yet to discover specificproof that any of the titled pien-hsiangbianxiang)recordedhereinwere used for purposesof narrative recitation. This is not, of course, to denythat they mayhave been so used, and, indeed, the likelihoodthat they were seems great. Still,in orderto avoid the imprecision concerningpien that has been rampant in the past, extremecaution should be exercised in making any claims about the relatednessofpien-wenandpien-hsiang."See ibid., p. 43.


6/83

WHAT IS BIANXIANG? 115and techniques were available to the pien-hsiang artist for indicating sequential ac-tions and events, his chief aim was generally to portray a story of some sort."5

Mair's reservations towards Bai's functional definition of bian-xiang, however, seem to have largely disappeared in his more recentpublications, as when he states: "Essentially, it was the task of theartist to represent these manifestations [of the Buddhist figures]on paper or silk, or in wall-paintings, in which case they may becalled pien-hsiang ("transformation scenes or tableaux"). The pien-storyteller would then use the pien-hsiangas an illustrative device dur-ing his performance.'s The two authors seem to have reached abasic consensus.

Thus, for more than half a century there has been a continuouseffort to establish a direct link between Dunhuang bianwen (andother types of literature) with paintings, including Dunhuangmurals. In this process scholars have gradually changed their focusfrom speculating on the relationship between bianwen and certain il-lustrations to hypothesizing that all paintings called bianxiangare il-lustrations of bianwen. This second argument, first introduced byhistorians of Chinese literature, has further influenced art historiansin their interpretations of Dunhuang art. For example, Shi Wei-xiang -Tt1 interpreted the Dunhuang murals of the Xianyujing VjM, or the Sutra of the Wise and theFoolish, as visual aids of "popu-lar suitra lectures" held in cave-temples.'7 Similarly, RoderickWhitfield commented on a fragment of a Dunhuang silk painting il-lustrating the contest between Raudraksa * g3- and Sariputra -I]1A:"This painting is of great interest as an example of the literarygenre known as pian-wen [bianwen],dramatized recitations by monksin return for donations from the faithful. They correspond to thepien-hsiang or pictorial illustrations of the sfitras, of which most ofthe wall paintings in the caves, as well as those on silk, are ex-amples. "1

15 Ibid., p. 43.16 Victor Mair, Painting and Performance: hinesePictureRecitationand Its Indian Genesis(Honolulu: University of Hawaii Press, 1988), p. 1.17 Shi Weixiang, "Guanyu Dunhuang Mogaoku neirong zonglu" IN_t# r%- PS 4f, in Dunhuang wenwu yanjiusuo #W:C 9 , ed., DunhuangMogaoku eirongonglu#M-Ajgp4M. (Beijing: Wenwu chubanshe, 1982), p. 195.18 Roderick Whitfield, TheArtof CentralAsia:TheSteinCollectionn theBritishMuseum, vols.(Kodansha, 1982), 1:318.


7/83

116 WU HUNGARE BIANXIANG "VISUAL AIDS" FOR BIANWEN PERFORMANCES?

The specific relationship between bianwenand bianxiangdefined bythese authors is of ultimate importance to the study of Chinese arthistory because it concerns the form, content, and function of the en-tire range of religious (and perhaps also secular) art from the fifthcentury to twelfth century. Unlike the term bianwen, which mayhave only designated a particular literary form and rarely appearsin historical records, the word bianxiangwas widely used during thisperiod for all sorts of Buddhist art and even for some non-Buddhistart works. The astonishing popularity of bianxiang s evident; it wassaid that the monk Shandao -4 (617-81) "painted more than threehundred walls with Pure Land bianxiang,'"9 nd that the late Tangpainter Fan Qiong l'$ covered more than two hundred walls withvarious bianxiang pictures.20 More bianxiang murals and picturescrolls are recorded in Tang-Song painting catalogues and otherwritings, and still more such paintings and engravings exist in Bud-dhist cave chapels. Based on these records, the modern scholarGuan Dedong W94 concluded, "It can be said that no Buddhisttemple during the Tang was without bianxiang."'Bai Huawen's and Mair's arguments reviewed earlier consist oftwo interrelated theses: (1) bianxiangwere narrative representations,and (2) bianxiang were used as visual aids in bianwen storytelling.A closer look at the historical concept of bianxiang and a carefulexamination of this type of painting, however, leads to radicallydifferent conclusions: (1) strictly speaking, most paintings labeled asbianxiangare not "narrative" in either content or format, and (2)bianxiangpaintings executed in Buddhist cave-temples were not usedin storytelling. Starting from the term bianxiangitself, we find thatits meaning seems to have changed over time. A major change oc-curred some time around the High Tang. Before the eighth century,

19 TakakusuJunjir6 AMi10&J3 and Watanabe Kaigyoku . r L,, ed., TaishJ hinshuiDaizokyo7EiSA)ZOR, 100 vols. (The Taisho issai-ky6 kanko kai, 1922-34) [hereafterT], 2035.39.365b.20 Huang Xiufu R4f, Yizhouminghuaud )i$l (Beijing: Renmin Meishu chuban-she, 1964), A.4.

21 Guan Dedong, "Tan 'bianwen' " =j 'VZ', originally appearedinJuequn houbaoVf01 1.1-12 (1946); L WL 1:199.


8/83

WHAT IS BIANXIANG? 117this term was used for various kinds of art forms, including three-dimensional statues, relief carvings, book illustrations, scroll paint-ings, and murals. An early reference to bian or bianxiang is found inthe Foguoji ({4n`, Record of Buddhist kingdoms), in which FaxianME recorded a Ceylonese ritual he witnessed in 410: "The king ex-hibits, so as to line both sides of the road, the five hundred differentbodily forms in which the Bodhisattva has in the course of hishistory appeared: -here as Sudana gAk, there as Syama(bian) f,now as the king of elephants, and then as a stag or a horse. All thesefigures are brightly colored and grandly executed, looking as if theywere alive. '22 The various forms of the bodhisattva described in thispassage were represented by painted life-size statues. This usage ofthe term bian or bianxiangwas confirmed by Duan Chengshi &tA,who wrote that "during the Liang dynasty (502-57) sculptured im-ages were also called bian."23Another type of bian-sculpture are miniature carvings. It wasreported, for example, that the Eastern Jin ruler Yao Xing 04t(r. 394-416) presented "various bianxiang" (zabianxiang* ) fromKucha to the Buddhist Master Huiyuan 8*24 These works are de-scribed as being executed in the form of "xilii , a term which

22 Based on James Legge's translation, A Record f BuddhisticKingdoms,Beingan Account ytheChineseMonkFa-hienof His TravelsnIndiaand CeylonA.D. 399-414), in Search f theBuddhistBooksofDiscipline Oxford: Claredon Press, 1886), p. 106.23 Duan Chengshi, Suxiangjif 2, in Yao Xuan tOZ, ed., T'angwencui I9;.Y (SPTKedition, Shanghai: Shangwu yinshuguan, 1929), 76.509a. An even earliertext, Luoyangjialanji 4MM-l2, records that Huisheng hired some artisans to make Sakyamuni's four stiupabian. See Wang I-tung, trans., A Record fBuddhistMonasteriesn LoyangPrinceton: PrincetonUniversity Press, 1984), p. 243. See also Mair, "Transformationtableaux," pp. 15-16.24 Huijiao 1I, Liang gaoseng huanRs*JW, in T 2959.360. Huiyuan's biography istranslatedby E. Ziircher in TheBuddhistConquestf China,2 vols. (Leiden: E. J. Brill, 1959),1:240-253. The identification of Yao Xing's gift as bianxiang,however, was challenged byMair, who argued that the three characters, zabianxiang, hould be read as "various,""unusual," and "shapes." See "Transformation tableaux," pp. 6-7. But similar termsdefinitely referring to bianxiang cenes are also found in painting cataloguesand include zafo-bianM N (various Buddhistbian)andzawu bianxiang2JV 1 (bianxiangf various things).See Pei Xiaoyuan -9*0, Zhenguan ongsihuashiJR9A 1, in Chen Liantang F11,ed., Tangdai ongshuShanghai: Jinzhang shuju, 1921), 58:22a., and Zhang Yanyuan 1,ILidai minghuai , in William Reynolds Acker, trans., SomeT'ang and Pre-T'angTextson Chinese ainting,2 vols. (Leiden: E. J. Brill, 1954-74), 2:203. Based on this evidenceand a possible relationship between Yao Xing's zabianxiang nd miniature carvings fromChinese Turkistan, I followZiircher's identificationof these carvingsas "various scenes fromscriptures executed in fine embroidery." See Ziircher, 1:249.


9/83

118 WU HUNGZuircher translated as "fine embroidery" and Mair interpreted as"fine thread" or "inlaid stone. "25The character lu is often used in-terchangeably with lou , meaning "to carve"; and the word xiloug can be explained as "fine carving" or "openwork carving."Buddhist art in the Gandharan area and Chinese Turkistan isfamous for a kind of miniature sculpture intricately engraved withscenes of the Buddha's life.26 They seem to fit the description of"fine carvings of various bianxiang" perfectly.

The word bianxiang,moreover, was used in the mid-sixth centuryto refer to two-dimensional relief decorations on a stutpa. YangXuanzhi *VJ1z5*ecorded that at Gandhara the traveling monkHuisheng MML hired skilled artists to copy in bronze a certain"Four-stuipa bian of Sakyamuni" E Y .27 As scholars haveargued, the iconography of this work must have resembled a stuipain the King Asoka Monastery 1 in Zhejiang, which the Tangmonk Jianzhen &A described as follows: "On one side is theMahasattva bian *EE9-, on another side is the Sunetra bian 6;iS, on another side is the Candraprabha bian A YJCE6T,nd onanother side is the Sibi bian _JtIE. X28 These four jataka scenes allappear in pre-Tang murals at Dunhuang.During the pre-Tang and early Tang periods, the term biandesignated not only statues and reliefs but also paintings. It isrecorded that the celebrated Liang dynasty painter Zhang Sengyou3M 0, was a master of jingbian S , or "suttra bianxiang.29 While itis not clear whether Zhang painted his bian-pictures on walls or silkscrolls, all six bian from the Sui royal art collection, catalogued bythe early Tang art historian Pei Xiaoyuan, appear to be scroll paint-ings.30The last usage of the term bian during this period was as tex-tual illustration. The "Bibliography" chapter in the History of SuiDynasty (Sui shu P4; compiled 629-36) lists two books, one about

25 Mair, "Transformation tableaux," p. 7.26 Some examples of such miniature carvings can be found in the exhibition catalogue,Afuganistanudaibijutsuen 'If*t,I#c (Nihon keizai shinbunsha, 1963), pl. 53-84.27 See Mair, "Transformation tableaux," p. 15.28 T 2089.989. The text only explicitly identifiesone scene: the bianof Mahasattva. Here Ihave identifiedthe other three pictures according to their descriptions: "giving up of eyes"(Sunetra), "giving away the head" (Candraprabha),and "the rescue of the dove" (Sibi).29 Yao Silian AT,1 Liangshu(Beijing: Zhonghua shuju, 1973), 54.793.30 See Mair, "Transformation tableaux," pp. 25-26.


10/83

WHAT IS BIANXIANG? 119horse-riding and the other about the game of pot-pitching, both con-sisting of a textual section and a bian (pictorial) section.3"From the High Tang on, the terms bian and bianxiang wereused more strictly: they no longer designated sculptured forms butreferred only to pictorial images; they were mainly associatedwith complex suitra illustrations and never with individual icons,even painted ones. We reached this conclusion in part by investigat-ing inscriptions that commemorated patrons who had Buddhistchapels constructed and decorated; in the four documents sum-marized below, the use of the term bian and bianxiang shows theseprinciples:

Cave 148 (776):2(non-bianxiang): sculptures representing a parinirvana scene,and Cintamanicakra a and Amoghapasa 7; painted icons of a Thousand Buddhas.(bianxiang): large murals of the Baoen , bian, the Tian-qingwen i bian, the Ma-njusri iZ* bian, the Saman-tabhadra 5- bian, the Eastern Bhaisajyaguru ~Titi bian,the Western Paradise Jti$? bian, the Thousand-armedand Thousand-eyed Avokokitesvara +F+R1f bian, the

Maitreya 0 bian, the Cintamanicakra bian, and theAmoghapasa bian.Cave 231 (839):"3' These are the QimadugeOU6 t and Touhujing&'.. Mair has argued that the termbian here simply means various ways of riding horses or playing the game of pitch-pot("Transformation tableaux," pp. 5-6), but the combination of ajing E (a Classic or a suitra)and a bianseems to indicate that such bian-picturesare the illustrations of the texts. See ZhouYiliang M A, "Du 'Tangdai sujiang kao' " *fl;tt , in LWL 1:163.32 This information is derived from "The Record of the merit achieved by his honor LiTaibin ; a native of Longxi "E during the Tang," which is inscribed on a stele inDunhuang Cave 148. He Shizhe :tt has compared the statues and paintings mentionedin the text with those in the actual cave. "Cong gongyangren tijikan Mogaoku bufen dongkude yingjian niandai" ,*A* in Dunhuangyanjiuyuan#W 5P, ed., DunhuangMogaokuongyangreniji ftMR-A #ig, A%W (Bei-jing: Wenwu chubanshe, 1986), pp. 205-6. The inscription is partially translated in Mair,"Transformation tableaux," p. 11.33 This information is derived from "The Record of the merit achieved under the Tibetansby the retired scholar Yin g at the Grottoes of Mogao Caves in the formerDunhuang prefec-ture." He Shizhe has compared the statues and paintings mentioned in the text with those inCave 231 at Dunhuang. See ibid., pp. 207-8.


11/83

120 WU HUNG(non-bianxiang): statues of Sakyamuni Buddha and bodhisatt-

vas; painted icons of Manijusri, Samantabhadra, and heaven-ly kings.

(bianxiang): large murals of the Western Paradise bian, theLotus TERbian, the Tianqingwen bian, the Baoen bian, theBhaisajyaguru bian, the Huayan #* bian, the Maitreya bian,and the Vimalakirti bian.

Cave 192 (876):'(non-bianxiang): seven statues centered on Amitabha RJprMWBuddha, as well as those of Ma-njusri and Samantabhadra;painted icons of Cintamanicakra, (Amoghapasa), the fourdirectional Buddhas, and 66 other Buddhist icons.(bianxiang): large murals of the Bhaisajyaguru bianxiang, theTianqingwen bianxiang, the Amitabha bianxiang, and theMaitreya bianxiang.

Dunhuang manuscript (S. 4860v) "A record of merit achieved"(Gongdeji 34MNE):3(non-bianxiang): statues of Sakyamuni Buddha and attend-ants; painted icons of Thousand-armed and Thousand-eyedAvolokitesvara, Cintamanicakra, Amoghapasa, four heaven-ly kings and attendants.

(bianxiang): large murals of the "Subjugation" bianxiang andother unnamed bianxiang.

These documents also include a number of measure words thatare regularly connected with certain types of art work: pu, "layout"or "composition," appears to be a general measure word, most fre-quently used for a group of statues or a complex bianxiangmural;while the word quV, or "image," is used for counting individualicons, either sculptured or painted. This distinction helps clarify aprevious confusion. Commenting on the record of "one pu of sevenstatues (centered on) the Amitabha Buddha in the (central) niche"of Cave 192, Mair states, "From this inscription, we gain the sureknowledge that a single 'layout' or 'spread' (p'u) could depict more

3T Fhis nscriptionin Cave 192 by the patrons Zhu Zaijing k:if:, Cao Shanseng land 28 othersis transcribed n ibid., pp, 84-85, and is brieflydiscussedby Mair in "Transfor-mation tableaux," p. 12.35 This text is translatedin ibid., pp. 7-10.


12/83

WHAT IS BIANXIANG? 121than one event or episode."36 But neither this passage nor themeasure word pu implies any narrative representation. In fact,these statues, which still partially exist in the cave, show a layout ofa static Buddha icon flanked by his attendants.

The claim that, from the High Tang on, the terms bian and bian-xiang pertained only to complex pictorial images is further supportedby several important painting catalogues and texts compiled duringthe period from the ninth to eleventh centuries. These includeZhang Yanyuan's (ca. 847-74) Recordof FamousPaintings of SuccessiveDynasties, Duan Chengshi's Record of Monasteriesand Pagodas (Sitaji+4nd, mid-ninth century), Zhu Jingxuan's *IA Recordof FamousPaintings of the Tang (TangchaominghuialuAVEf, mid-ninth cen-tury), Huang Xiufu's FamousPaintings of the Yizhou(Yizhou minghualu, ca. 1006), and Guo Ruoxu's B Record of Paintings Seen andHeard (Tuhuajianwen zhi F Q.r.X, , late eleventh century). No lessthan forty-four bianxiangare catalogued, and they are all paintings,including both temple murals and picture scrolls.37Moreover, theirtitles and descriptions suggest that they were complex compositions,not individual icons.

Information on the meaning of bianxiang during and after theHigh Tang may also be derived from divergent textual sourcesdating from the eighth century and after. Most of these materialshave been collected by Mair and will not be cited here.38 Again, thebianxiangmentioned in these documents are mostly paintings. Oneexception is a nirvana bian engraved on a stele, made for EmpressWu A in the Dayun Monastery k;+;39 but this scene must have

36 Mair's understandingis partially influencedby his misinterpretationof the charactershi$, whichhe translatesas "event." See ibid., p. 12. Here this charactermeans "work" or "apiece of work," as we find in other Tang writings. See CiyuanO, (Beijing: Shangwu yin-shuguan, 1988), 66. The phraseyipu qishi -"- , thus refersto a group of sculpturescon-sisting of seven statues.37 Zhang Yanyuan records at least 44 bianxiang f 19 kinds, painted by famous artists insplendid Buddhist temples in Chang'an and Luoyang. Duan Chengshi records 6 bianxiang f5 kinds. Zhu Jingxuan twice mentions bianxiang.Huang Xiufu records 25 bianxiang f 17kinds. Guo Ruoxu records 8 kinds of bianxiang.These passages are collected in Mair,"Transformation tableaux," pp. 26-39.38 Mair, "Transformation tableaux," pp. 16-25.39 See ibid., p. 17. Xiang Da [ mentions a "nirvana bian" in the Longmen Caves, but,as Mair suspected, this may in fact refer to the stele.


13/83

122 WU HUNGbeen executed in low relief or in line engraving, and is thus again atwo-dimensional representation."

From studying these sources, we have arrived at the conclusionthat, from the High Tang on, bianxiangwere generally understoodas a kind of two-dimensional, complex pictorial representation.Because they are two-dimensional, they are not sculptures,4' andbecause they are complex pictorial representations, they exclude in-dividual icons. Yet the term bianxiangis certainly not synonymouswith "painting" or "pictorial art"; it only refers to certain typesof religious, mostly Buddhist, painting. In fact, except for a fewreferences to Daoist paintings,42 the huge amount of works labeledas bian or bianxiang n the above sources are strictly Buddhist. Not asingle secular painting created during the Tang-Song period wascalled bianxiang.

This observation forces us to realize a fundamental discrepancybetween bianxiangand bianwen.The two most frequently quoted tex-tual references to the bianwen performance-two poems by Ji Shilaoand Li He $ -both suggest that the performers of bianwenwerelay people-a singing girl in one case and a concubine in the other.43Bianwen may be defined in many different ways but can never beclaimed as a "religious" literary genre.44According to Mair, seven

40 A similarstele is discussed by AlexanderC. Soper, "A T'ang parinirvanoatele," ArtibusAsiae12.1/2 (1959): 159-69.41 This is also suggested by Duan Chengshi's remark, "During the Liang dynastysculpturedimages were also called bian." By "also" Duan implied that during his time bianno longer referredto sculptured images.42 Emperor Zhongzong 42 of the Tang declared that he heard that Daoist followers

throughoutthe country had all been painting the bianxiangf Laozi's conversionof theBar-

barians. See Mair, "Transformation tableaux," p. 41. Guo Ruoxu mentioned a bianxiangfa Daoist scriptureby Feng Qing j and the XuanhehuapuVi g of the Song containsthe record of another Daoist bian by Dong Boren J{It. See Mair, "Transformationtableaux," pp. 38, 40.43 For Englishtranslationsof the two poems, see Mair, Tang Transformationexts,pp. 152,155. For other textual evidence for the secular nature of bianwenperformances, see GaoGuofan AMW "Lun Dunhuang minjian bianwen""'OWUJt, in GansushengshehuiKexueyuan wenxue yanjiusuo : ed., Dunhuangxueunji#

NAV (Lanzhou: Gansu renmin chubanshe, 1985), pp. 188-94; and Mair, T'angTransforma-tionTexts,pp. 152-60.44 See n. 1. It should be noted that Mair has made a similar distinction: "The data ad-duced in this article indicate thatpien-hsiangwere largely productsof the elite (includingtheecclesiastical establishment). This is in contrast to pien-wenand its oral antecedents whichwere situatedin the folkand popularrealms." See Mair, "Transformation tableaux," p. 43.


14/83

WHAT IS BIANXIANG? 123Dunhuang texts can be securely identified as bianwen.45Of these,five are historical, secular stories, and two are based on Buddhistsuitras. Not coincidentally, only these two have their counterpartbianxiangrecorded in Tang painting catalogues,46 and only the bian-xiang of the "Subjugation of demons" appears in the DunhuangCaves.

The heading of the "Bianwen on Mahamaudgalyayana' s rescue ofhis mother from hell" originally included a phrase "with pictures"(bingtu 1;Ji), which was later crossed out, perhaps because the il-lustrations were lost. Another Dunhuang text, "Story of HanQinhu 40Y," ends with the sentence, "The illustration (huabena*) has come to an end."'7 In Ji Shilao and Li He's poetic descrip-tions of performances of the "Zhaojun bian," pictorial illustrationsare introduced as huajuanI (picture scroll) or shuzhi , (Sichuanpaper).48 It should thus be clear that all pictures directly related tobianwen, as well as the only extant definable bianwenillustration ofthe "Subjugation of demons" (Fig. 5), are painted scrolls, notmurals. It should also be clear that these actual bianwen llustrationswere never called bianxiang.We must consider why in an age whenthe term bianxiangwas so frequently used to describe painted scenes,the writers or copiers of these bianwenscripts seem to have deliberate-ly avoided this word.

A general principle in studying religious art, including Buddhistcave-paintings, is that individual pictures and statues must be ob-served in their architectural and religious contexts: they are notBut in my opinion, it is difficult o classifybianxiangs "elite art," since such paintingsin Bud-dist caves and temples were still created for the masses.

45 These are "Bianwen on Mahamaudgalyayana's rescue of his mother from hell,""Bianwenon the subjugation of demons," "Bianwenon the Han general Wang Ling Ti, ""Wang Zhaojun bianwen,""Li Ling 2 bianwen,""Zhan Yichao Y fi bianwen,"and"Zhang Huishen 3iti bianwen."See Mair, Tang Transformationexts, pp. 17-23.

4 Fu Yunzi mentioned a "Mahamaudgalyayanabianxiang" LWL 1:154), but I have beenunable to locate the source. A "Subjugation" bianxiangwas also executed by two religiouspainters, Li Yongji ,R. and Li Xiangkun , in the Anguo %g Temple. SeeAlexander C. Soper, KuoJo-shi's Experiencesn Painting:An EleventhCenturyHistoryof ChinesePaintingTogether iththe ChineseText n Facsimile Washington, D. C.: American Council ofLearned Societies, 1951), pp. 52, 98; cited in Mair, "Transformationtableaux," p. 39.47 For discussions of this case, see Mair, Tang Transformationexts,pp. 11-12.48 See n. 43.


15/83

124 WU HUNGportable objects that can be carried around and appreciatedindependently but are components of a larger pictorial programdesigned for a particular ritual structure for religious worship. Inother words, Dunhuang murals were an integral part of cave-temples and the act of painting (and viewing) was itself a ritual prac-tice. Our interpretation of such "works of art" must take thesepoints into account.Applying this idea to analyze the bianxiangpictures in Dunhuangcaves, we find without much difficulty that it would have been im-possible to use these paintings for bianwen storytelling. The subjectmatter of surviving bianwen stories, after all, was predominantlysecular and the original performers and audience were in most caseslay-people. Unlike the picture scrolls used in bianwenperformances,Dunhuang bianxiangare integral parts of cave temples, which fallinto four architectural types of distinctive religious functions.Those belonging to Type 1 have a number of tiny cells attached tothe main chamber (Fig. la), which contains a large Buddha statueat the rear and is richly decorated with iconic and narrative sceneson its walls and ceiling. In the best preserved cave of this type, Cave285, the walls are covered by images of Buddhas, bodhisattvas,various Buddhist deities, and meditating monks, as well as some ofthe most beautifully executed narrative scenes from the WesternWei period (535-56). The largest narrative sequence appears on thesouth wall and illustrates the story of five hundred bandits who, sym-bolized by five figures, appear six times in the picture: fighting withsoldiers, being punished by officials, wandering in the wilderness,devoting themselves to Buddhism, and attaining enlightenment. Thisand other compositions in this cave unmistakably fall into a typicalnarrative mode called a "sequential" or "cyclic" representation.50

49 See Xiao Mo _, "Dunhuang Mogaoku de dongkuxingzhi" 4t',A9nl _02mti,in Dunhuang yanjiusuo, ed., Zhongguohiku:DunhuangMogaoku F f 5vols. (Beijing and Tokyo: Wenwu chubanshe and Heibonsha, 1982-87), 2:187-199. Xiao'stypology also includes the "Nirvana cave" and the "Giant Buddhacave," but caves of thesetypes are few and not germane to my study.50 In Kurt Weitzmann's definition, a "cyclic" narrative is "a series of consecutive com-positions with separateand centeredactions, repeating the actorsin each and so observing atthe same time the rulesof the unity of time and place." See his IllustrationsnRollandCodex:AStudy ftheOrigin ndMethod f Text llustration Princeton:Princeton University Press, 1947),p. 17.


16/83

125

ab

dyasy fro Duhagmgou129Ib ae24(ye2,Nrhr e

dynasty; ibid. 1:208. (c) Cave 45 (Type 3), High Tang; ibid. 3:239. (d) Cave 196(Type 4), late Tang; ibid. 4:236.


17/83

126 WU HUNGHowever, they were definitely not used in popular story-recitation,because this cave was built for the practice of meditation by monks.This type of cave, called a Vihara(monastery) in Sanskrit, origi-nated in India. The tiny cells, where monks sat in silence for daysand nights, flanked the main chamber; the Buddha statue in thismain chamber provided them with the subject of mental concen-tration, while the narrative pictures presented them with precedentsof devotion. The religious idea and practice associated with theViharacaves at Dunhuang are entirely alien to any kind of secularentertainment.

Caves belonging to Type 2 have a sculptured "pagoda" in thecenter (Fig. lb) and again derived their basic structure from an In-dian prototype. The religious practice related to this architecturalform was commonly known as circumambulation, in which the wor-shiper, entering the cave by the east gateway, walked around thepagoda in a clockwise direction. Since the central pagoda (which de-veloped from the Indian stuipa) was a chief symbol of Sakyamuni,circumambulation expressed the devotion to the historical Buddha.In practicing this ritual the worshipers accumulated good deeds. Sopreached the Suitraon theLife and Conductof theBodhisattva Pusa ben-xing jing t4i*eY): "Those who keep circumambulating aroundthe Buddha's [image] or the place of a Buddha's pagoda will gaincountless blessings."' It is interesting to note that many narrativepaintings created during the Northern Dynasties, including jataka-tale pictures and illustrations of Sakyamuni's life, appear in suchpagoda-caves (where ten of the eleven compositions of the Buddha'slife scenes are found)." Bai Huawen has interpreted these picturesas visual aids in bianwenstorytelling." But considering the religiousfunction and symbolism of the pagoda-caves, we can only concludethat these murals, whose major theme is the Buddha's own accu-mulation of merits extending back into countless rebirths in thepast, functioned to inspire faith among the worshipers. Such a closerelationship between pictorial narrative and the ritual of circum-ambulation had been displayed in Indian Buddhist art as earlyas the second and first centuries B.C.; stories from the Buddha's

51 Quoted in Xiao Mo, p. 190.52 These are Dunhuang Caves 254, 260, 263, 428, 431, and 290.53 Bai Huawen,p. 505.


18/83

WHAT IS BIANXIANG? 127lives were illustrated on the gates and railings along the circularpaths surrounding the great stuipas at Safichi and Bharhut.54 It isdifficult to imagine that in these Buddhist holy sites the ritual of cir-cumambulation could be replaced by or mixed with colorful story-telling for public amusement. If the narrative scenes were "visualaids," they could only have been assisting the Buddhist rite.Both the Viharaand pagoda caves were popular before the Suiunification of China. They have close ties with Buddhist architec-ture of India and Central Asia and an ideology that has strongHinayana overtones, as pictures in these caves often stress monasticpractices, self-discipline, and aloofness from society. In contrast,other types of Dunhuang caves have an architectural style that isessentially "Chinese," and their murals and sculptures, which il-lustrate Mahayana su-tras, attest to the dominance of Mahayanadoctrines. A majority of Sui and Tang caves belong to Type 3 (Fig.lc), while most caves built during the Five Dynasties and Song fallinto Type 4 (Fig. ld). Scholars have demonstrated that these newcaves imitated the wooden-framed Buddha or Image Hall in amonastery, which in turn was derived from the Throne Hall in theimperial palace.55 In cave Types 3 and 4, the central pagoda hasdisappeared, leaving a Buddha statue as the main icon, either in alarge niche on the rear wall (Type 3), or on a U-shaped platform inthe middle of a cave (Type 4). The form of religious worship asso-ciated with these cave chapels and their interior decorations wasguanxiangOW or "contemplating holy icons."A number of Mahayana suitras teach the guanxiang exercise,through which one aims to visualize the true images of variousBuddhas and bodhisattvas.56 Often by initially concentrating on apainted or sculptured icon, the worshiper could attempt "a sys-tematic building-up of visual images (in his mind's eye), each as

54 SeeJohn H. Marshall and AlfredC. A. Foucher, TheMonumentsfSdiichi, 3 vols. (Lon-don: Probsthain, 1940);Alexander Cunningham, TheStupaof BhdrhutLondon: W. H. Allenand Co., 1879).5 Xiao Mo, pp. 194, 197.56 For an introduction to these texts and the practice of guangxiang, ee Stanley K. Abe,

"Art and practicein a fifth centuryChinese Buddhist cave temple," Ars Orientalis0 (1991);see also Arthur Waley, A Catalogue of Paintings RecoveredfromDunhuang by Sir Aurel Stein (Lon-don: British Museum, 1931), pp. xii-xiii; AlexanderC. Soper, Literary videncefor arlyBud-dhistArt in China Ascona: Artibus Asiae, 1959).


19/83

128 WU HUNGcomplete and precise as possible, in a sequence from the simpletoward the complex "57 t is possible that guanxiangtexts did not orig-inate from India but were invented by Central Asian or ChineseBuddhists.58 These texts had appeared by the mid-fifth century, butit was during the Tang dynasty that the guanxiang practice becameespecially popular, as indicated by the appearance of Chinese com-mentaries on these siutras and abundant jingbian (sutra bianxiang)related to the ritual. The monk Shandao, who painted some threehundred Amitabha bianxiang n his lifetime, also composed an essay,"Methods for the merit of samddhiby contemplating the image ofthe Amitabha Buddha" (Guannian Amituofo xianghai sanmei gpngdefarnen ,?;1 rEfo a m&i ),in which he instructed:"Ifthere is a person who paints and constructs a bianxiangof the adorn-ments of the Pure Land in accordance with the Suitraon VisualizingAmitdbha and contemplates the divine place day and night, . . . Ifthere is a person who paints a bian-picture in accordance with thesuitra and contemplates its adornments such as the divine trees,divine ponds, and divine buildings, he will absolve his multitudi-nous sins in his present life. "59Such demands that religious devotionbe expressed by dedicating and contemplating "sutra bianxiang"explain the great proliferation of such art works at Dunhuang indi-cated in Table 1.

57 Soper, Literary vidence,p. 144.58 See Abe, "Art and practice. ' The guanxiang racticewas also associatedwith some earlyChinese Buddhistsects. One of the most importantevents in the historyof Chinese Buddhismwas the establishment of the Pure Land Sect by Huiyuan: in 402, this master monk assem-bled his followerson Mt. Lu and together they mnadehe vow to be rebornin the Buddha'sWestern Paradise.The vow was made beforean image of Amitabha and accompaniedby anoffering of incense and flowers. As Ziircher has remarked, "This urge to have a concreteobject of worship, perceptible by the senses, characterizes the Buddhism of the Lushan.Everywhere, in the biographical records as well as in Hui-yiian's own writings, we findthe same stress on visual representation: the use of icons in meditation, visualization ofAmitabha, his hymns to the 'shadowof the Buddha,' the dharmakhya transcendentbody) ofthe Buddha and[that of the Bodhisattvaetc." Zfircher, 1:220.59 T 1959.25.60 These statistics are based on the motif index of flunhuang nurals provided in Dun-huangwenwuyanjiusuo,ed., Neirong onglu,pp, 221-42. I excludeillustrationsof theParinirvin.a

Suitra,he Xianyuing or the Sutraof the WiseandtheFoolish,and the Vitnarlakirtiiitra.TheTang Parinirvd4abianxiangollows earlier Nirvainascenes depicting an episode in the life ofthe Buddha and does not show an exact "iconic" representation; the Xianyuing, which isessentiallya collectionof Buddhisttales, is not a typical "Sfitra"; while Vimalarldrtibianxiangare composed according to the "oppositional" mode and will be discussedlater. The method


20/83

WHAT IS BIANXIANG? 129TABLE 1: Dunhuang "suitrabianxiang"Bianxiang Pre- Tang Post- TotalTang early high mid late TangAmitabha rt 1 12 7 7 12 24 63Amitayus KM'X 0 1 21 34 18 10 84Maitreya At 5 6 14 24 17 21 87Bhaisajya il 4 1 3 21 31 26 96Lotus XTE 2 1 5 7 8 13 36Avalokitesvara 0 0 4 2 2 5 13Huayan ffi0 0 1 5 9 14 29Baoen VERM 0 0 2 7 11 12 32Tianqingwen )IP 0 0 1 10 8 13 32Jin'gang a0U 0 0 0 8 9 0 17Jinguangming *; E 0 0 0 4 4 3 11Lengjia 4bRN 0 0 0 2 5 5 12Siyi Fantian wen T X r1 0 0 0 1 3 8 12Miyan MON: 0 0 0 0 2 2 4Fumu enzhong 3 0 0 0 0 1 0 1

It should be noted that all the bianxiang listed in Table 1 are"iconic" representations, not typical pictorial narratives." Differ-ing from a narrative painting, whose main purpose is to tell a story,an iconic depiction is a symmetrical composition centered on anicon62- a Buddha or a bodhisattva. The visual centralization is notonly apparent in the icon's extraordinary size and solemnity, but isalso reinforced by surrounding figures and the architectural setting,which guide the viewer's gaze toward the central icon. The mostof dividing the Tang dynasty into four sub-periods is commonly accepted by Dunhuangscholars, but the exact duration of each period is defined differently. In Shi Weixiang's opin-ion (which is also the "official opinion" of the Dunhuang Academy), the four periods are:early Tang (618-704), high Tang (705-80), mid-Tang (Tibetan period) (781-847), and lateTang (848-906). See Shi Weixiang, p. 177. Osvald Siren, however, proposes anotherperiodization: early Tang (618-712), high Tang (713-65), mid-Tang (766-820), and lateTang (821-60). See Siren, Chinese ainting:LeadingMasters nd Principles, vols. (New York:Ronald Press, 1956), 1:86.

61 For a discussion of "iconic" and "narrative" representations in early Chinese art, seeWu Hung, The Wu Liang Shrine, heIdeology f Early Chinese ictorialArt (Stanford: StanfordUniversity Press, 1989), pp. 132-36.62 In this study, instead of followinga modern semioticdefinition, I use the term "icon" todesignate a sacredimage, specificallya type of painting that intended to be the focus of prayerand devotion.


21/83

130 WU HUNGcrucial difference between this type of mural and a narrativerepresentation, however, lies in the relationship between the pictureand the spectator. In a narrative painting the principal figures arealways engaged in certain events, acting and reacting to oneanother. The composition is thus essentially self-contained; and thesignificance of the representation is shown in its own pictorial con-text. The viewer is a witness, not a participant. In an iconic scene,the central icon, portrayed frontally as a solemn image of majesty,ignores the surrounding crowds and stares at the viewer outsidethe picture. The composition is thus not self-contained; althoughthe icon exists in the pictorial context within the composition, itssignificance relies on the presence of a viewer or worshiper outsideit. In fact, the openness of the composition is based on the assump-tion that there is a worshiper who is eng-aged,in direct relationshipwith the icon. It is based on this assumption that the iconic composi-tion has become universal in various religious art traditions aroundthe world.63Given their ritual functions and pictorial forms, the iconic "suitrabianxiang" in the Dunhuang caves could not have been used instorytelling. It is true that in some "suftra bianxiang,"such as theLotus bianxiang and the Amitayus bianxiang, the central icon is en-circled or flanked by "narrative scenes." These scenes, however,are based on selected passages from various chapters of the sutrasand rarely form a coherent narrative sequence.64 Sutras, such as the

63 For example, we find that the EasternChristianchurchdecreedthat icons partakeof thespiritualessence of the figuresthey depict and that they constitute "the essential point of con-tact betweenthe human and divine realms." See G B. Ladner, "Concept of the image in theGreek Fathersand the Byzantine iconoclastic controversy," DumbartonOaksPapers7(1953):10.

64 For example, in a typical Lotus Sutra bianxiang,he Buddha image is surrounded byscenes thatderive their subjectsfrom the twenty-sevenchaptersof the siutra.See Shi PingtingPERrnd He Shizhe, "Dunhuang bihua zhong de Fahuajingbiang chutan" # R bt,&* WV , in DunhuangMogaoku,3:171-91; Joseph L. Davidson, The LotusSuitranChinese rt (New Haven: Yale University Press, 1954); Matsumoto Eiichi * Tonko-ga nokenkyu 3D7Rr,, 2 vols. (T6h6 bunka gakuin T6ky6 kenkyiijo, 1937), 1:110-42.An Amitayus bianxiangften includes a depiction of Ajatasatru's story in a side panel. But asArthurWaley has pointed out, this illustration is most likely based on Shandao's commen-tary on the Sustra,ratherthan illustrating the original script. See An IntroductionotheStudy fChinese ainting New York:Grove Press, 1923), p. 128. See also Matsumoto Eiichi, 1:45-59.Instead of telling stories, these so-called "narrative" scenes allude to and explain Buddha'steachingand arethus similarto the "analogies" and "parables"used widely in sutra ectures.


22/83

WHAT IS BIANXIANG? 131Huayanjing (Buddhavatamsakamahavaipulya-su-tra) and the Yaoshiliuliguang jing gMO (Bhaisafraguruvaidaryaprabha-szitra),rehighly abstract, and their bianxiang either contain no narrativescenes or include discrete illustrations of the "twelve great vows ofthe Bhaisajya Buddha" and the "nine forms of violent death."65Again, at Dunhuang there are numerous bodhisattva bianxiangwhose sole content are the images of these Buddhist divinities andtheir attendants.66 Because of their boundless compassion, thesebodhisattvas were believed to be able to intervene in people's livesto save them from danger, help them to achieve happiness, andguide them to bliss in the afterlife. The fundamental method of seek-ing help from these compassionate deities was to invoke theirnames, recite their merits, and reproduce and contemplate their im-ages. At Dunhuang these images were copied from large muralsinto woodblock prints. Among these is an image of Avolokitesvaraaccompanied by instructions on the proper use of the icon:He who wishes by reciting this spell to obtain the protection of the Holy One mustfirst in a clean place put this holy image, make appropriate offerings, and then sub-mit and do obeisance before he recites. He must with whole heart accept and doobeisance to all the Tathagatas, put away all turbid emotions and share the natureof the Great Merciful Arya Avalokite?vara Mahasattva; and he must pray that allSentient Beings may be in this accord, himself doing obeisance ten times with headand visage. Then he must sit straight and with undistracted heart concentrate uponthe recitation of Arya Avalokite?vara's Lotus Flower Division Heart-spell, to wit:Om Arolika. Svaha. The might and virtue of this heart-spell are wide and great.It can diminish evil-karma and keep away disaster. It can prolong life, increase pros-perity. If you can recite it a full 300,000 times, even the most evil karma can becompletely abolished, and no misfortunes or disasters can molest you. You canbecome intelligent and eloquent to any extent you please. If you can manage torecite it a million times all sentient beings who behold you will attain a completelyillumined state of mind and in the future certainly be born in Paradise, to such1;v-;+a na nirm - ^n ;,n thoa r%r;frr;al aii+vn 67

65 See Waley, Catalogue,pp. 62-70.66 In addition to the many Guanyin jing . bianxiangthat are centered onAvalokitesvara,at Dunhuang there are 65 Cintamanicakra bianxiang, 7 Amoghapasa bian-xiang, 40 Thousand-armed and Thousand-eyed Avalokitesvara bianxiang,10 Water-moonAvalokitesvarabianxiang, 31 Maiijusri bianxiang, 6 Qianshoubo f ifk Maiijusrl bianxiang,and 125 Samantabhadra bianxiang.These statistics are based on Dunhuang wenwu yan-jiusuo, ed., Neirongzonglu,pp. 221-42.67 Trans. in Waley, Catalogue,p. 195-96. The woodblockprint is reproduced n Whitfield,2:140.


23/83

132 WU HUNGThe chief compositional mode of Dunhuang art was, from thevery beginning, "iconic." All 492 Dunhuang caves, built over a

period of more than seven hundred years, contain large or smallicons in painted and sculptured forms, and iconic murals constituteover ninety-five percent of the total pictorial compositions. It iswell-known that depictions of jataka tales (bensheng*4), Buddhistdidactic stories (biyu -llk; avaddna),and the Buddha's life (benxing*fY) flourished during the Northern Dynasties and the Sui. But evenduring this "golden age" of Dunhuang narrative art such paintingswere limited in number: in the 110 caves dated to this period onlytwenty-seven pictures illustrating eleven jataka tales have beenfound. Depictions of Buddhist didactic tales are even fewer-sevenillustrations of five stories.68The narrative content and format of these paintings, however, donot automatically suggest that they were used in oral storytelling.Our question is: Could they have been practicallyused in bianwenperformances? The Dunhuang caves are carved into a mountaincliff. Most of them consist of double chambers connected by a nar-row corridor, while others have a single chamber with a passage-way sometimes thirty feet long. In front of each cave was built awooden-framed facade.69 With the caves designed in this way, therear chambers, which contain most paintings and statues, wouldhave been very dark. Even today, when most wooden facades havecollapsed, a visitor can barely make out the tiny images and inscrip-tions in these cave-chambers. Rather strangely, the ceiling of acave, which is especially obscure, is often richly decorated with nar-rative scenes. In fact, typical narrative paintings in the "sequen-

68 These jataka tales include the Syamajataka (Caves 461, 438, 299, 301, 302), the Rurujataka (tl,l, Cave 257), the Sibi jataka (Caves 275, 254, 302), the Brahmana rjataka (caves 285, 302), the Candraprabhajataka(Caves 275, 302), the Sunetrajataka (Cave302), the Mahasattva jataka (Caves 254, 428, 299, 301, 302, 419, 417), the Sujata Wjataka (Cave 296), the Sudanajataka (Caves 428, 419 423, 427, 423, 427), the Kalyanakarin$:t:? jataka(Cave 296), the Pradyotapradipa MEgEataka (Cave 302), the Byilin-gar-li11MMMEE Ataka(Caves 275, 302), and theJalavahana WL*AA jataka (Cave 417). Bud-dhist didactic tales indude the story about a suicidal monk (Caves 257, 285), Nanda's Urstory (Cave 254), Sumati's r*XtC story (Cave 257), the story of five hundred bandits(Caves 285, 296), and the storyabouta virtuous nun (Cave 296). See Takada Osamu A FOiS,"Fojiao gushihua yu Dunhuang bihua" : in DunhuangMogaoku,2:200-208.


24/83

WHAT IS BIANXIANG? 133tial" style created during the Northern Wei and Sui dynastiesalways appear on the ceiling.70The difficulty in seeing these cave-paintings disappears when westudy the pictures in beautifully printed catalogues, which are nowthe chief research source of those scholars outside the DunhuangAcademy. When individual scenes and statues are photographedwith the help of modern technology and reproduced with extraor-dinary clarity, they are divorced from their contexts and seemanalogous to individual "paintings." They are readily viewed asself-contained "pictorial stories," or as "visual devices" in certainoral performances. The most telling example of this kind of mis-reading concerns murals depicting the story of the "Subjugationof demons," which various authors have used as evidence for therelationship between Dunhuang wall-paintings and bianwenperfor-mance. Cheng Yizhong, for example, making much of the word chuA (the place where . . .) in the bianwenversion of the story found inthe secret library at Dunhuang, argued that the storytellers reliedon the mural:We find that most Dunhuang illustrations of this bianwenare centered on Sariputraand Raudraksa-the two main characters of the story-and that various episodesin their magical contest are woven between these two figures to form a coherent car-toon-like pictorial representation. If the "place" (chu) of each episode were notspecifically indicated in an oral recitation, the audience would not be able to com-prehend the picture. Thus the story-teller pointed out explicitly the "place" (chu)of an episode in every verse part, so the listeners would find the correspondingscene in the mural. This is a definite proof of how bianwenand bianxiangwere closelyused together.7"

Cheng's theory sounds convincing, especially when we find in thenewly published GrottoArt of China: theMogao Cavesat Dunhuangthatthese murals are full of dramatic details painted in brilliant colorsand inscribed with explanatory cartouches often copied from thebianwen.However, this theory seems to work onlywhen these muralsare studied from these reproductions; in the actual caves they canhardly be seen, and would be extremely difficult to be used in a

69 Xiao Mo, p. 188.70 These outstanding narrative paintings can be found on the ceilings of Caves 290, 296,301, 303, 423, 420, and 419. See Takada Osamu, p. 208.71 Cheng Yizhong, p. 389.


25/83

134 WU HUNGperformance. The kind of mural Cheng mentioned appeared duringthe late Tang and continued to enjoy popularity through the Song.The most elaborate of these murals are found in Type 4 caves, inwhich a group of statues stand on a U-shaped platform in front of alarge partition (Fig. Id). A pictorial representation of the story ofthe "Subjugation of demons" often appears on the rear wallsbehind the central platform, with a narrow space about three feetwide between it and the partition. As two art historians, Li Yong-ning 7*8 and Cai Weitang AM*, have pointed out, it would beimpossible for a picture at that particular location in a dark cave-chamber to serve any kind of oral recitation.72

At this point, we might ask whether bianxiangmurals were de-signed to be viewed in a conventional manner at all. The disciplineof modern art history is largely based on the theories of perception:analyses of forms and perspective always assume a viewer throughwhose eyes the researcher observes and interprets a work. We cantrace this notion to the fifteenth-century Italian theoretician Alberti,who first defined "painting." According to Alberti, as summarizedby Svetlana Alpers, "[a painting] is conceived of as a window ontoa second world. The viewer, rather than the world seen, has prior-ity."73 It is no accident that this theory did not appear until theRenaissance, when painting began to gain an independent statusand was commissioned by individual patrons.If a painting was not made to be seen in this "modern" manner,our whole methodology based on viewing and reading is shaken.We are forced to answer the question: who was the supposed vieweror reader? It is said that a master sculptor and his student were mak-ing an icon. While the master was carefully modeling and decorat-ing the back of the statue, the student became impatient and askedhim, "Why do you waste paint and time on a spot no one willsee?" To this his teacher answered, "The god will see." This anec-

72 Li Yongning and Cai Weitang, " 'Xiangmo bianwen' yu Dunhuang bihua zhong de'Laoducha doushengbian' " *,t7M M, in Dunhuang wenwuyanjiusuo, ed., 1983 nian quanguoDunhuang xueshu taolunhui wenji (Shiku. Yishu bian)-LJ_

+ISITJ2gXJ8n T (Eg, R4iV), 2 vols. (Lanzhou: Gansu renmin chubanshe,1985), 1:187-88.73 SvetlanaAlpers, "Art history and its exclusions:the example of Dutch art," in FeminismandArtHistory,ed. Norma Broudeand Mary D. Garrard(New York: Harper&Row, 1982),

p. 185.


26/83

WHAT IS BIANXIANG? 135dote reveals two fundamentally different attitudes toward artisticcreation. To t-he student, the purpose of the statue was to please or-dinary viewers like himself. To the master, the act of making thestatue was an expression of religious commitment. He did not ruleout the problem of "viewing," but, for him, the viewers includednot only worshipers but also the unseen god, and the value of hiswork lay primarily in his devotion to the god.This anecdote thus reveals an important aspect of religious artthat is often overlooked in a standard artistic analysis. Some seventhousand inscriptions in the 492 caves at Dunhuang explain the pur-poses of making caves and images. They were written by donors ofdifferent social statuses from local rulers to a peasant wife, and theyvary in length and literary style. But the central theme of these in-scriptions remains the same: devotion. In a thorough check of theseinscriptions, I have found not one word pertaining to "seeing" or"exhibiting" (or, for that matter, to "performance"). Most shortinscriptions follow a simple formula: "[the donor] dedicates," or"[the donor] dedicates wholeheartedly." Long inscriptions, whichare often entitled "a record of merits and virtues" (gongdeji), oftenbegin with praise for the Buddha and the greatness of his benevo-lence. They then describe in detail the process of constructing thecave, the statues and murals, and the donor(s)' hope that the meritearned from this devotion will be transferred to the ruler, localcommunity, and deceased ancestors or living relatives.74 Short orlong, these inscriptions document the donors' hopes that by con-structing a cave or painting a bianxiang hey could accumulate merit.The same idea must have motivated people to make caves and im-ages on vertical cliffs and high ceilings. The near impossibility ofreaching these dangerous places became itself a challenge in reli-gious devotion and artistic creation.

Returning to the murals depicting the "Subjugation of demons"at Dunhuang, these paintings present another problem that haslong troubled their researchers. Following the commonly acceptedapproach that these works are illustrations of the bianwen, a re-searcher naturally begins his investigation by identifying painted

74 Such long inscriptions include the "records of merit and virtue" of Zhai Fengda (642,Cave 220), Li Huairang a (698), Li Taibin (776), a gentlemanYin, Wu sengtong% ,Suo Falu , and the Zhai N family.


27/83

136 WU HUNGscenes (and their inscriptions) with certain passages in the literarysource; he then arranges the scenes sequentially, according to thetextual narrative. The result of such a research is shown in Fig. 2,in which all scenes in a "Subjugation" mural have been identifiedand are replaced by serial numbers. In terms of iconography suchresearch seems fruitful, but exactly at this point the researcherreaches a dead end. The sequence indicated in the painting by hisnumbering system is illogical because it does not suggest any "se-quence" or "order" of visual images. To read these scenes accord-ing to their assigned numbers, it is necessary to shift our gazefrom one corner to another, to cross the whole width of the painting(which could be forty feet across), or to "scan" the complex com-position to search for a minute detail. Our eyes and mind would spinuntil we got totally dizzy and finally gave up. However, all "Sub-jugation" murals created during the late Tang and Song have suchirregular compositions.

The design of the "Subjugation" murals might be explained inseveral ways. It is possible that, since the "Subjugation" story had

(D I _ ___~~ ? ( @00 0

Fig. 2. Reconstruction of the narrative sequence of the "Subjugation" bianxiangin Dunhuang Cave 9; from Li Yongning and Cai Weitang, p. 173.


28/83

WHAT IS BIANXIANG? 137become so well-known by the late Tang, visitors to the caves wouldfind no difficulty in reading these illustrations no matter how theywere organized. It is also possible that such a painting was adeliberate "pictorial riddle," which demanded deciphering andthus engaged interest. One might also argue that because of thedifficulty in comprehending the mural, a storyteller was necessaryin guiding the audience through the painted scene.75None of these ex-planations, however, are persuasive enough to resolve the problem.If it was indeed assumed that the contemporary visitors could easilyidentify the painted scenes, the explanatory cartouches would havebeen unnecessary. If the mural was a deliberate "riddle," thensome quick clues would be evident, but instead these pictures seemonly to demand hard labor. And if the mural was designed for astoryteller conducting a tour, then surely it would have been locatedwhere there would be enough room. Stymied by the dilemma, somescholars have finally found a solution: the design of these murals issimply a "mystery."To resolve this and other problems we need new angles of observa-tion. I want to propose that (1) devotional art is essentially an artof image-making rather than image-viewing, and (2) the process ofimage-making has its own logic that differs from those found inwriting and oral recitation. Based on these two propositions I willclosely analyze a large group of murals depicting the tale of the"Subjugation of demons," whose various compositions createdover the course of five hundred years provide us with excellent ma-terials in studying the nature of Dunhuang bianxiang, especially itsrelationship with Dunhuang literature.

7 See Cheng Yizhong, p. 389. Indeed, scholars have noticed that picturesused in storytell-ing, such as the painted bannersand scrolls in the par traditionin India, do not alwaysfollowa linear narrative sequence. Joseph C. Miller, for example, has observed that "Some pic-torial scenescombined two or more narrated events in a visually integratedway. This findingunderscoresthe different communication potentials between linearly patterned verbal nar-rativeand spatiallypatternedpictorial narrative." See J. C. Miller,Jr., "Current Investiga-tions in the Genre of Rajasthani par painting recitations," in Winand M. Callewaert, ed.,EarlyHindiDevotional iteraturen Current esearch,Orientalia ovaniensia nalecta o. 8, p. 118.But sincepaintingsused in these traditionsare portable scrollsand banners, they arenot com-parablewith Dunhuang "Subjugation" murals and cannot be taken as direct evidence in in-terpreting the function of these murals.


29/83

138 WU HUNGBIANXIANG AND BIANWEN: DUNHUANG PAINTINGS

ON THE "SUBJUGATION OF DEMONS"Painting and literature on the "Subjugation of demons" have

been well researched. Various versions of the story have been foundin a number of sutras and in a complete bianwentext,76 and we alsoknow at least twenty-one Dunhuang paintings (eighteen murals, apicture scroll, a fragment of a silk banner, and a set of ink drawings)depicting the tale.77 The bianwen text has been studied by manyhistorians of Tang literature and has been excellently translated intoEnglish by Mair, 78and the paintings have been catalogued and dis-cussed by art historians in China, Japan, and the West.79 Whilethese studies have shed much light on the development of this impor-

76 Altogether four copies of the Xiangmobian% , were found in the secret library atDunhuang. One, which originally belonged to Luo Zhenyu HI, is now missing and onlya transcriptof the text survivesin Luo's book, Dunhuangingshi#W . Anotherversion, inthe Pelliot collection (P. 4615), is broken into six pieces and severely damaged. The thirdmanuscript (S. 4398) is also a segment and contains only the first forty-one lines of thebianwen.Fortunately, the fourthcopy appearsto be complete. The two fragmentsof this ver-sion are preserved in the British Library(S. 5511) and the Beijing Library.77 Li Yongning and Cai Weitang listed nineteen Dunhuang murals depicting the "Sub-jugation" story. The one in Cave 94, however, is identified accordingto a text. The pictureitself can no longer be seen. See Li Yongning and Cai Weitang, pp. 170-71. For scholarlyworks on these paintings, see n. 79 below.78 Mair, Tun-huangopularNarrativesCambridge:CambridgeUniversityPress, 1983), pp.31-84.7 Waley, Catalogue, o. LXII; Nicole Vandier-Nicolas, Sariputrat lessix maitres 'erreur.Facsimile du Manuscrit Chinois 4524 de la Bibliotheque Nationale. Mission Pelliot en AsieCentrale, Seriein-Quarto, V (Paris: ImprimerieNationale, 1954);Matsumoto Eiichi, "Ton-k6chih6ni ryuik6 eshiR6takusha t6sei henso" YMP 5t~LA 1 AtrM . , Bukkydgeijutsu1 #f 19 (1923); "R6takusha t6sei hens6 no ichi danpen" 1iY, Kenchikuhi7WSP 2-5 (1940); and Tonko-gaokenkyu, p. 201-1 1. Akiyama TerukazutKIIi5h, "Tonk6-bon g6ma-hen (R6takusha t6sei hen) gakan ni tsuite" JV4**A

(2 tg 1CO 'C, Bijutsukenkya F57 ' 187 (1956): 1-35; "Tonka ni okeruhenbun to kaiga-Futatabi R6takusha t6sei-hen (g6ma-hen) o chulshinni" t llfW:k L M, Bijutsukenkyu 11 (1960): 1-28; and "R6takusha t6seihen hakuby6funpon"l R H4MV ** (Pelliot Tibetain 1923), Tokyd aigaku ungakubuunka oryuz enkyuihise-tsukenkyuayo ; nos. 2 and 3, 1978;Jinweinuo *f2l, "Dunhuang bihua Chiyuanjitu kao" ftQg;jf* ft, WenwuCankao iliao ;7.iV W, 1958(8): 8-13, rpt. in LWL1:341-52; and "Chiyuanjitu yu bianwen" 1A=xtKUr, in ibid., 1958 (11): 32-35, rpt. in LWL 1:353-60; Luo Zongtao 5gW,* "Xiangmobianwen huajuan" 1X? F, in Zhongguoudianxiaoshuoyanjiuhuanji 'pJffit RFt;gt (Taibei: Taibei shuju, 1979); Jao Tsong-yi , et.al., PeinturesMonochromeseDunhuangParis: Ecole FrangaiseD'extreme-orient, 1978); Whitfield; Li Yongning and CaiWeitane.


30/83

WHAT IS BIANXIANG? 139tant theme, there is a general tendency to emphasize a one-sidedrelationship between art and literature, in which pictures passively"illustrate" the bianwen.80 would argue, however, that the relation-ship between the paintings and texts was far more complex anddynamic. Most noticeably, the development of visual imageryfollowed its own logic. The story was constantly restructured andenriched in pictorial forms, and new images in turn influencedliterary compositions and contributed to the general development ofthe story. A close examination of such interactions will not onlyreshape the developmental sequence of this specific tale, but willalso sharpen our method in studying Dunhuang art and literaturein general.

The full version of the "Subjugation" tale consists of two looselyconnected parts. The first part concerns finding the Buddhist holysite of Jetavana Garden 1fUNI.n his travels to the kingdom of Ra-jagrha E to arrange his son's marriage, Sudatta M-, a ministerof the kingdom of Sravasti ; heard that the Buddha was anenlightened man. He paid tribute to the Buddha, and upon receiv-ing the Buddha's instruction he immediately became a devotedBuddhist. He hoped to invite the Buddha to preach in his nativekingdom, but for this he first had to build a monastery in a perfectlocation. Thereupon the Buddha sent his disciple, S?riputra, toaccompany and guide Sudatta in his search for a site for themonastery. After some time they found a garden that belonged tothe Crown Prince of Sravasti. Sudatta covered the garden's groundwith gold-the price for the site-and dedicated the place to Bud-dhism.

The second part of the story begins with an abrupt turn: it has anew theme-the magical contest between Buddhists and heretics-and new central characters-the two competitors Sariputra andRaudraksa. The news that Sudatta was going to build a Buddhistmonastery in Jetavana Garden had reached the ears of the heretics.Confident of their magical power, they proposed to the king thatbefore a monastery could be established, the Buddhists would haveto compete with them. The contest was witnessed by the king and

80 It should be pointed out, however, that in their article on the "Subjugation" pictures,Li Yongning and Cai Weitang have noted many differences between these murals and thebianwen tory. But as I have shown earlier, their identification of these picturesstill follows theliterary narrative and neglects the paintings' own structure.


31/83

140 WU HUNGall his officials, and six times in a row Sariputra demonstrated hissuperior power over Raudraksa, the chief magician of the heretics.The story ends with the surrender and conversion of the hereticsand, upon the Buddha's arrival, the whole kingdom embraced Bud-dhism as the only truth.This summary is based on an early version of the "Subjugation"story, entitled "Sudatta builds a monastery" i in theXianyujing, a collection of Buddhist tales recorded by eight Chinesemonks at a Buddhist convention in Yudian.8" After being assembledin Turfan, the book was taken in 435 to Liangzhou in Gansu, wherea monk named Huilang MM assigned it the present title. Evidencesuggests that the far more elaborate Dunhuang "Subjugation"bianwenwritten three centuries later was based on this sutra.82Mosttellingly, the bianwen version follows the bipartite narrative struc-ture of the Xianyu jing story, while other pre-Tang versions of thetale contain only the first part of the tale.83

Interestingly, the earliest depiction of the "Subjugation" tale atDunhuang is also based on the Xianyujing. This picture (Fig. 3) isfound in Cave 12 in the Western Caves of a Thousand Buddhas #foA1iand has been dated to the Northern Zhou.84 The compositionis roughly divided into two registers each containing a number ofinscribed scenes of unequal width. The narrative proceeds from leftto right on the upper register, and then turns back on the lowerregister. Although partially blackened by smoke, its twelve scenesand accompanying inscriptions are still generally comprehensible.

81 T 202.418-22.82 A detail in the bianwen ives the text a definite date: in the "prelude" EmperorXuan-zong A'_/Vs referredto by a lengthy title-"Emperor of Kaiyuan and Tianbao who is a Sagein Civil Affairsand Divine in Military Matters and who Responds to the True Way" Mi3tt"St#;X;F- ZhengZhenduoirstnoticed hatXuanzongassumedhisroyal itle n748but soonchanged t in 749. See KennethK. S. Ch'en,Buddhismn China: HistoricalSurveyPrinceton:PrincetonUniversityPress, 1964), p. 289; Li Yongningand CaiWeitang,p.169.83 ThesesiitrasncludeTanguo* (2ndcentury),Zhongbenqijing:uda in pEf. A

i r; TanWuchandE (385-433),Daban iepanjing:hizihoupusapin;.@&M. 0I9fLtM '; TanWuchan,Fosuoxingzan:uageiGudupin JiJ. {L*ND.lrJ; andHuiyanffi (5th century):Daban iepaning:ShizihouusapinIAAMA. a-T RGrlr

84 This cave s assignedheno. 10 in oldcatalogues.Themostup-to-datentroductionothiscave s inDunhuang anjiuyuan,d., Chzigokuekkutsu:nseiyurinkutsu jMtA Heibonsha,1990),pp. 291-93.


32/83

141

0bIo

_ ~~~~~~~~~~~~~~~~~~~~~~~~nc

C14

10

t ~~~~~~~z- , br.

\JNJ _ u

-~4b


33/83

142 WU HUNG(Upper register):Scene 1. In this largest scene in the painting, the Buddha sits infront of a building surrounded by trees. Sariputra stands beside theBuddha while Sudatta is bowing to them. The cartouche reads:"Before Elder Sudatta returns to Sravasti to build the monastery hebids farewell to the Buddha. The Buddha and Sariputra together[commission] the construction of the monastery. [This is] the mo-ment when [Sudatta] says good-by to the Buddha."

Scene 2. Separated by a tall tree from the previous scene, twofigures walking toward the right are identified in an inscription:"Elder Sudatta and Sariputra go to Sravasti to build the monasteryfor the Buddha . . . [the rest of the inscription is blurred]."Scene 3. Sariputra sits inside a building. The rest of the pictureand the inscription are unclear.

Scene 4. Sudatta and Sariputra converse among lush trees.Although the cartouche is illegible, the picture may represent theirfinding Jetavana Garden.85(Lower register):Scene 5. A large tiger devours a water-buffalo. The cartouchestates, "[This is] the moment when Raudraksa transforms himselfinto a water-buffalo, and Sariputra transforms himself into a lion."

Scene 6. The picture and inscription are completely blackened.Scene 7. Sariputra receives homage from a kneeling figure, prob-ably a demon. This and the previous scene may together representone of the six magical contests between Sariputra and Raudraksa,in which the holy monk transforms himself into a white elephantwhich dries up a pond conjured up by the heretic. The inscription iscompletely blackened.Scene 8. A large bird catches a dragon. The inscription explains,"Raudraksa transforms himself into a dragon, and Sariputra trans-forms himself into a gold-winged bird."Scene 9. An image of a leaning tree. The inscription reads,"[This is] the moment when Raudraksa transforms himself into abig tree, and Sariputra transforms himself into a whirlwind to blowit down."

85 Jin Weinuo has suggested that this scene depicts the episode in which Sudatta informsSariputraabout the forthcomingcontest (LWL 1:344). But this episode is not included in theXianyujingversion and is only found in later bianwen.


34/83

WHAT IS BIANXIANG? 143Scene 10. Although the accompanying cartouche is blurred, thescene clearly depicts the contest in which Sariputra transformshimself into the Diamond Warrier to destroy Raudraksa's moun-

tain.Scene I1. This scene is placed to the left of the first scene on the

upper register." The inscription explains, "[This is] the momentwhen Raudraksa transforms himself into a monster and Sariputratransforms himself into Vaisravana Maharaja sitbimkTE, who[then] burns the monster with fire. '

Scene 12. Raudraksa throws himself on the ground in front ofSariputra. This scene at the lower left corner of the compositionends the pictorial narrative.These are solid grounds for identifying the source of this mural asthe Xianyu jing story. As just mentioned, this satra was the onlysixth-century text to include both Sudatta's search for JetavanaGarden and the contest between Sariputra and Raudraksa. Oncloser examination, it also appears that some episodes illustrated inthe mural are recorded in the suitra but not in the later bianwen.Most obviously, the contest between King Vaisravana and themonster is described differently in the satra and the bianwen,and themural closely illustrates the following passage in the satra:Again, he (i.e., Raudraksa) transformed his body into a yaksa demon. His size wasenormous; fire shot from his head. His eyes were as red as blood; his four teethwere long and sharp. With flames issuing from his mouth, he bounded forward.Then Sariputra changed himself into the Vaisravana Maharaja. The yaksa wasterrified and wanted to retreat at once. Fire sprang up on all four sides so there was

86 The ratherconfusing position of this scene has ledJin Weinuo to identify it as the firstscene of the whole pictorial narrative(L WL1:344-45). In my opinion, however, the irregularplacement of this scene was very likely accidental: the painter did not first sketch out thewhole composition,but ratherpaintedscene by scene as he went along. Towardthe middle ofthe lower registerhe would have alreadyrealized that there might not be enough spaceleft forthe whole narrative. So scenes are increasinglysqueezed together and cartoucheswritten onincreasinglynarrowerspaces. Even then, not enough spaceremainedon the lower registerforthe last two scenes. The artist thereforeraised the contest scene to an empty space left on theupper register (where he actuallypainted this scene over the trees surroundingthe buildingbehind the Buddha), so he could still paint the last scene representing Sariputra'sfinalvictoryat the lower left corner to conclude the whole composition.

87 In his article on the Dunhuang "Subjugation" paintings Akiyama misread JinWeinuo's article and quoted the textual reference from the Xianyuing as the inscription ofthis scene.


35/83

144 WU HUNGno place to escape. Only on Sariputra's side it was cool and there was no fire. Theyaksa submitted right away by throwing himself on the ground in an attitude of pro-found reverence and begging plaintively that his life be spared. As soon as he feltshame, the fire disappeared. The crowd cried out in unison, "Sariputra is the win-ner Raudraksa is no match for him "88The bianwenoffers a very different account of what happened whenthe heretic conjured up two monsters: "Sariputra hesitated while heconsidered what to do. Before long, Vaisravana appeared in front ofthe King with a leap. His awe-inspiring majesty was splendiferous,and his armour and weapons were bright and shiny. Earth deitiessupported his feet; and a precious sword hung at his waist. This iswhere the two monsters, after taking one look at him, incessantlybegged for their lives.""

Two major differences between the siutra and the bianwen arerelated to the iconography of the Cave 12 mural. First, in the suitrathe monster surrenders because he is burned by fire and finds noway to escape, but in the bianwen the two monsters are instantlydefeated "after taking one look at" the heavenly king. Second, inthe suitra Sriputra and Raudraksa "transform themselves" intothe divine king and the monster, respectively; in the bianwen theheretic "conjured up" the two monsters while King Vaisravanaoffered his own service to the holy monk. The mural inscription(which specifies, "Raudraksa transforms himself into a monsterand Sariputra transforms himself into Vaisravan.a Maharaja whoburns the monster with fire") is apparently based on the satra.Moreover, the last two scenes in the painting actually follow the nar-rative sequence of the sutra: in Scene 11 the monster is burned byfire, and in Scene 12 Raudraksa throws himself on the groundbefore the victorious Sariputra.

Although the plot of the story is somewhat rearranged in themural,90 an important parallel between the pictorial and literaryrenresentations is found in their overall narrative structure. The

88 Mair, T'ang TransformationTexts, pp. 54-55, trans. slightly modified.89 Mair, PopularNarratives,p. 80-81, trans. slightlymodified. To save space I have placedMair's translation of the prose section into a continuous paragraph.90 Some important iconographicdiscrepanciescan be found between the mural and thetext. For example, the pictorial narrativeof the six contests follows an order different fromthat in the text:


36/83

WHAT IS BIANXIANG? 145composition of the picture mirrors the "bipartite" division of thetext: all scenes on the upper register are related to Sudatta's searchfor the holy site, while those on the lower register focus on themagical contests. The painting thus closely follows the literary, tem-poral narrative; and because many details of the story are omitted,the layout of the mural brings the basic structure of the text into asharper focus.

Some researchers have proposed that the painting's "sequential"format originated in Han dynasty pictorial carvings, which weredelineated in horizontal bands.9" But to my knowledge all knownHan narrative paintings (including bas-reliefs) are "episodic,"meaning that a story is always represented by a single scene.92It hasalso been suggested that the narrative scheme of this mural wasderived from a type of Chinese narrative hand-scroll painting thatemerged during the Wei-Jin period, represented by Gu Kaizhi's af1iL "Nymph of the Luo River" PVMM.73 It is true that often in ahand-scroll painting various episodes of a story are presented in alinear fashion and accompanied by paragraphs of a text. But a

Xianyu ing Cave 10 mural(1) tree-whirlwind (1) buffalo-lion(2) pond-white elephant (2) pond-white elephant(3) mountain-Diamond Warrior (3) dragon-gold-winged bird(4) tree-whirlwind (4) dragon-gold-winged bird(5) mountain-Diamond Warrier (5) buffalo-lion(6) monster-King Vaisravania (6) monster-King VaisravanaJin Weinuo, who first studied this mural, proposed that this iconographic inconsistency"dearly demonstrates the beginning stage of the narrativerepresentationof the story, whichhad not achieved its maturity" (LWL 1:345). It seems to mhe,however, that the simplestmethodto illustratea text is to followits original sequence, not to alter it, and that the painterof the mural knew the story quite well, for the scenes on the upper registerobedientlyfollowthe narrative of the text. Thus the irregular sequence of the contest scenes on the lowerregistermay have resultedfrom a deliberateeffort to increase the dramatic effectof the storyby rearrangingthe events. The new pictorial sequence seems to reflect the painter's concernwith the gradualheightening of the drama: during the contest Siiriputraassumes increasinglypowerfulforms(in the Buddhisthierarchy) froma lion to a white elephantto the gold-wingedbird (garuda)to a whirlwind (Wind God) to the Diamond Warrior(vajrapani) and finally tothe divine king Vaisravana. In fact, as I will discuss later, in the development of the "Sub-jugation" tale and its visual representations the contest sequence was subject to constantchanges.91 Jin Weinuo, in LWL 1:345.92 See Wu Hung, pp. 133-34.

93 Ibid.


37/83

146 WU HUNGhand-scroll must be read from right to left and the story presentedmust proceed in the same direction. This sequence differs from theCave 12 mural, in which the story develops from left to right on theupper register and then reverses direction on the lower register.More likely, this format originated in Indian Buddhist art-such asGandhara relief carvings, in which individual scenes of a story aresometimes arranged from left to right.94When this structure was in-troduced to China, it was enriched by pictorial devices developedthere. In the "Subjugation" mural, not only are the figurativetypes, buildings, and trees clearly derived from Chinese pictorialvocabulary, but landscape elements are used here to divide thecomposition into individual compartments, just as they are in theEastern Jin brick reliefs portraying the Seven Worthies in the Bam-boo Grove Yxt;W5Land the Northern Wei sarcophagus carvingsrepresenting the lives of filial sons.95

If this earliest "Subjugation" mural mimics literature in nar-rative structure, the second earliest depiction of the tale at Dun-huang (Fig. 4) rebels against literature by reconstructing the storyinto a new framework. The early Tang date (686) of this muralin Cave 335 is obtained from an inscription on the north wall ofthe cave. The west, or back, wall of the cave is occupied by a largerectangular niche, in which a statue of the Buddha was originallyflanked by his disciples and bodhisattvas. Rising from the rear wallof the niche to the ceiling are painted cloud patterns, and abovethe clouds appears the "jeweled pagoda" in which the Sakyamuniand Prabhfitaratna ** Buddhas reside. The two side walls of theniche are painted with a "Subjugation" bianxiang, a representationthat differs entirely from the one discussed above.The most radical change made by the early Tang artist is that hehas completely omitted the first part of the story about Sudatta'ssearch for Jetavana Garden. The mural is entirely devoted to the

94 I am indebted to Higashiyama Kengo 3IJa* who makes this point in his paper onillustrationsof the Syamajatakapresentedto the InternationalConferenceon Dunhuangolo-gy, Dunhuang, 1990.95 For illustrationsof these two works, see Yao Qian MU and Gu Bing t#, Liuchaoyishuo (Beijing: Wenwu chubanshe, 1981), pl. 162; Laurence Sickman and AlexanderC.Soper, TheArtandArchitecturef China Baltimore: Penguin Books, 1956), pl. 53.


38/83

147

N

~~~~~~~~~0bfD

bF~f3~~~~~~~~~~~C

-' - 0~~~~~~~bI,0

bb


39/83

148 WU HUNGmagical contest between Sariputra and Raudraksa. Therefore, itcan no longer be called " Sudatta builds a monastery, " after the titleof the story in the original Xianyu ing. More appropriate titles wouldbe the "Subjugation of demons" or "Raudraksa's contest with thesage"-the labels of murals recorded in Tang painting cataloguesand the titles of the eighth-century bianwen.96Because all existingtexts, both suitrasand the bianwen, include the first part of the story,we can reasonably assume that the new focus on the magical com-petition was established in the domain of visual art.

This new focus is closely related to the pictorial mode of the Cave335 mural, which I call an "oppositional composition." Unlikemost "suitra bianxiang," which are centered on a single Buddha orbodhisattva, paintings in this category show two juxtaposed figures,left and right. While these two main figures are engaged in a par-ticular event-a contest or a debate between them-other eventsfrom the same story are often depicted as secondary elements tofill in the picture. With such a structure, the mural of Cave 335divorces itself from the linear, temporal mode essential to theNorthern Zhou work. Images are now isolated into two separategroups: on the north wall of the niche Sariputra sits on a platformunder a canopy and is accompanied by Buddhist monks; on the op-posite south wall Raudraksa appears in the midst of male andfemale heretics. Scen

Documents

Bian Xiang