1. The Dunhuang Cave Discovery and Early Dispersal of Material
One of the richest discoveries for early medieval history of central and east Asia and their links to cultures beyond was made in June 1900 when a hidden chamber was uncovered at a Buddhist rock-cut temple complex near the town of Dunhuang, in what is now Gansu Province in the Peoples’ Republic of China (PRC). This small chamber, hereafter referred to as the library cave, contained tens of thousands of manuscripts, hundreds of paintings, textiles, artefacts and some printed documents. Dating from the fifth to the eleventh centuries, most texts were Buddhist scriptures in Literary Sinitic and Old Tibetan, reflecting the dominant powers in this region in this period. But there were also manuscripts showing Dunhuang’s links to the wider world, both geographical and cultural, among then Sanskrit manuscripts on palm leaf, as well as Khotanese, Old Uyghur, Sogdian and Judeo-Persian fragments (Galambos 2020: 13–17; van Schaik & Galambos 2012: 29–34).
No evidence has been found to confirm why this material was placed in this chamber. As van Schaik and Galambos discuss (2012: 18–28), various hypotheses have been offered, from it being a cache of sacred waste to being the contents of a library from a local Buddhist monastery awaiting conservation. The reason and date for its sealing and disguise are also conjecture: Rong notes the various scholars who have suggested that it might have been prompted by threats of invasion (1999: 254). Van Schaik and Galambos suggest that it might have been rather a considered act because the storage space was full and no longer of use (2012: 26–28). This is the theory originally offered by M. Aurel Stein (1862–1943), in 1907 the first foreign explorer on the scene following the cave’s discovery. This paper is based largely on Stein’s annotations on the manuscripts he acquired, his accounts of his 1907 visit published in 1912 and 1921 (the former a popular narrative and the latter a detailed report) and on his archives now in the Bodleian Library, Oxford (hereafter MSS Stein). The archives include his diaries (typewritten copy at MSS Stein 204–5) and pocket notebooks (MSS Stein 141–4), but also several other sources, such as letters to friends and colleagues and later official reports. Although many accounts have been written of his visit, culminating in the excellent work of Rong Xinjiang (1999 and 2013: 79–108), previous authors did not have access to all this documentation and thus their accounts are not comprehensive.
According to the reports we have, the cache was discovered by accident on 22–3 June 1900 by the resident self-appointed custodian of the cave, a monk from central China named Wang Yuanlu 王圓籙 (c.1849–1931), when workmen under his direction were clearing out sand from the corridor to another chamber. However, it is important to stress that there is no known direct documentary evidence for this account: it is based on later oral reports. Nor is there any known documentary or visual evidence of the chamber when first discovered. The earliest known photographs and written records of the cave are from Stein when he visited in May 1907.1 Despite Rong’s assertion that it was the visit of Stein that led to the disruption of the cave’s original order (2013: 115), there is compelling evidence to suggest that Wang had removed and replaced the cave’s contents well before Stein’s visit. It is also without doubt, as Stein mentioned, that Wang examined and then gave away material on several occasions from 1900 onwards (Stein 1912: 167–8; Fujieda 1966–70; Rong 2013).
Rong explores this in some detail, based mainly on the diary of a contemporary local official Ye Changchi 葉昌熾 (1849–1917), Gansu Provincial Commissioner of Education in Lanzhou, and on the writings of the painter Xie Zhiliu 謝稚柳 (1910–97), from the latter’s time at Dunhuang in 1942–3 (Xie 1949). Xie’s account records several dispersals of the manuscripts before Stein, starting from 1900 when Wang sent material to Yan Dong 延棟, an official based in Jiuquan east of Dunhuang. Yan Dong gave some to ‘a Belgian man’ from the tax office who, in turn, presented some to General Changgeng (長庚) (2013: 85).2 Recent research published here by Foong Ping tracks a Dunhuang manuscript fragment in the Seattle Art Museum collection back to an official, Rao Yingqi 饒應祺 (1837–1903), and the evidence suggests that he might have acquired the scroll from a ‘Belgian’, Paul Splingaerd (1842–1906), or from other people in Rao’s circle in Xinjiang and Gansu. Rong also explores also how other manuscripts remaining with Yan Dong were later dispersed, ending up in several private and public collections.
There were other recipients over the following years, including Ye Changchi, who was presented by Wang Zonghan 汪宗翰 (magistrate of Dunhuang from 1902–06), with four manuscripts, a painting and a rubbing in December 1903.3 In 1904, as detailed in his diary, Ye received more material, this time from another Dunhuang official, Wang Zonghai 王宗海, including leaves from a Sanskrit manuscript pothi (Rong 1999: 89–91).4 Rong reviews those items now in collections worldwide from this and other pre-1907 dispersals by Wang and although ‘we do not know for sure how much material came out of the cave during this early period’ (102), he suggests that it was hundreds of Chinese scrolls and paintings. He further suggests that some of this material, such as the paintings presented to Ye Changchi, must have been nearer the top of the cave’s contents because they were in excellent condition, that is, they had not been crushed by a weight of material piled on top of them for centuries (92). However, Monique Cohen (2002: 26–30) also discusses paintings now in the Freer Gallery of Art possibly from Ye Changchi and discusses the hypothesis that some are forgeries.
Of course, it is possible that Wang could simply have examined and given away some of the most accessible material rather than remove the cave’s contents in their entirety, but Rong’s appraisal that ‘in comparison with the contents of the entire cave library, which has been nearly completely published by now, some of the manuscripts and paintings removed by Abbot Wang are still among the best pieces’ (101) would suggest that Wang had looked through a considerable amount of material to find good examples. It would have been natural for him to remove all the contents on the cave’s initial discovery both to find out what it contained, as well as to uncover any wall paintings and statues, his main interest. This is corroborated by Stein who writes: ‘any indications that the original position of the bundles might have afforded at the time of discovery, had been completely effaced when the recess was cleared out, as the Tao-shih [Wang] admitted, to search for valuables, and again for the purpose of removing the big inscribed slab from the west wall into the passage outside.’ (1912: 182; 1921: 808–9). Stein noted the slab in the corridor during his 1907 visit: ‘The inscrn. now walled in on l. of passage was also first discovered in recess. The date is T’ai Ching 9 = 855 A.D.’ (May 23, MSS Stein 294/307).’ Stein’s Chinese assistant, Jiang Xiaowan 蔣孝琬 (d.1922), made a transcription, while Stein made a rubbing, and the text was later translated by the French scholar, Édouard Chavannes (1865–1918) and published in Stein’s expedition report (1921: Pl. CLXXV, Appendix A.).5 Given the dimensions of the cave, it would have been very difficult if not impossible for the stele to have been removed without the cave having been cleared of its contents.
Although Stein was not allowed to photograph the cave, Wang let him look in and he gave the following description: ‘Heaped up in layers, but without any order, there appeared … a solid mass of manuscript bundles rising to a height of nearly ten feet, and filling, as subsequent measurement showed, close on 500 cubic feet.’ (1912: 172).6 He further noted that those ‘bundles filled with miscellaneous text, painting fabrics, ex-votos, papers of all sorts.…by their irregular shape and fastening …could readily be distinguished from the uniform packets containing rolls of Buddhist texts in Chinese and Tibetan. Fortunately their very irregularity had caused the Tao-shih to put them on the top when he had built up the wall-like array of what I may call ‘library bundles.’ (182–3) He suggests that several such bundles had been placed at the very bottom of the cave by Wang ‘to turn a low clay platform into a level foundation for the manuscript wall above’ (190).7 They contained many paintings and some embroideries.
Apart from the initial disruption, dispersal and probable reordering of cave contents without any documentation prior to Stein’s visit, there are other complicating issues. First is the fact that, as noted by Stein, material was stored in other caves and some of this was almost certainly placed in Cave 17 after its discovery by Wang but no record had been made to distinguish this material from that found in Cave 17, or to note its original find site (1921: 828–9).8 This is corroborated by French Sinologist Paul Pelliot (1878–1945) During his 1908 visit he carried out a survey of other caves and found more manuscripts. Yet more were discovered in excavations by the Dunhuang Academy of the northern caves between 1988 and 1995 (Peng and Wang 2000–4).9 Pelliot also thought that some material had been placed in Cave 17 after its opening: ‘Pelliot…thinks there were two strata of mss. in the library of Tunhuang—the first consisting of the bundles, all of which were sealed up before the end of the 11th century, and the second a miscellaneous lot brought in from collections of the Mongol period, 14th–15th cent., and thrown loosely on top of the Ancient Fonds.’ (MSS Stein 65/28–9)10 Some of the material clearly dates from later.11
2. Stein’s Acquisitions and Documentation
A. The order/disorder of material in the Library Cave
We will first look at what we can tell of the ordering of the contents of the cave from Stein’s writings and photographs of the cave’s entrance and of various manuscript bundles during his visit in May–June 1907. The earliest known photographs of the interior of the cave were not by Stein but taken in 1908 by Charles Nouette (1869–1910), the photographer on Pelliot’s expedition. These show Pelliot sitting in the cave surrounded by manuscripts (Figure 1). The manuscripts appear to be largely in three formats: piles of flat items, marked A in Figure 1, are most probably formed from piles of largely Tibetan pothi; scrolls which are wrapped together in bundles, marked B, are what Stein called ‘regular library bundles’; finally, messier bundles, marked C in the picture.
Figure 1
Paul Pelliot examining manuscripts inside the library cave in 1908. Photograph: Charles Nouette. Musée Guimet AP8187, with annotation by authors.
Two questions immediately present themselves. First, when Stein arrived in 1907, were all the bundles in the same place as when the cave was opened in 1900? Secondly, did most of the bundles still contain the same manuscripts and paintings as when found in 1900?
Stein considered these questions. Regarding the first, he concluded that Wang had replaced the material in a different order: ‘The utter confusion prevailing in these bundles and their careless fastening, often without an outer cloth cover, clearly showed that no trouble had been taken to preserve the materials in whatever kind of arrangement they might have originally been found’12 (1921: 811). He noted that the ‘miscellaneous bundles’ were on top and concluded that Wang had left these last when replacing material in the cave after his initial discovery and emptying of the cave.
On the second question, which is key to our argument here, Stein describes ‘regular bundles’ wrapped in hemp covers, which were mostly sewn up, and considers whether these covers were original or added by Wang (1912: 190).13 He notes that Wang asserted the former. Several of Stein’s photographs shows piles of bundles (Photo 392/27(587–9) and see Figure 9), sewn but ‘the ends are generally left open.’ (1912: 190; 1921: 823). The wrappers shown in these photographs appear to be no more than pieces of cloth with little shaping while other scrolls have no outer wrapper.
Figure 9
Several regular bundles in their hemp covers, one with an inscription. British Library, Photo 392/27(589).
There is the additional issue of several more sophisticated purpose-made scroll wrappers which were among the finds. Such wrappers were usually made of several layers, with silk or hemp outers reinforced with layers of paper inside, often pieces cut from presumably discarded manuscripts. Some were made with a bamboo outer layer or only from paper. One end was sometimes triangular and had a fabric tie extending from its point to secure the wrapper around the scrolls. Others were rectangular and had ties both inside and out. Very few have survived in their original state: once it was discovered that they contained fragments of manuscripts in their structure, most were taken apart. Two exceptions of the finer type are MAS.858 (Ch.xlviii.001) and MAS.859 (Ch.xx.006), both in the British Museum and originally found in miscellaneous bundles by Stein. It is, of course, possible that they originally enclosed manuscripts removed by Wang and it is plausible that he would have focussed on these—their quality compared to the hemp wrappers might have suggested to him that they contained the finer scrolls.14
Wang Yuanlu had no shortage of time or opportunity to open all the bundles if he had wished. Given that, as discussed above, he gave away several hundred scrolls which were of good quality, he must have looked at a reasonable number. But we argue here that he left many untouched, based on Stein’s observation that many of the covers were still sewn up.15 And while he must have opened some, it is also probable, we argue, given the number of scrolls remaining in many bundles visible both from photographs and from Stein’s documentation, that he only removed one or two of the finest scrolls from these, if any.16 Since Wang’s knowledge and interest was with the Chinese material, it is also reasonable to conclude that he ignored most, if not all, of the Tibetan material.17
We therefore do not address here the issue of how the bundles of the manuscripts and paintings were originally placed in the cave but rather look at whether it is possible to reconstruct any of the individual bundles from the annotations Stein made on the manuscripts, based on the assumption that the contents of most of the regular bundles were as originally placed in the cave.
B. Stein’s acquisitions from the Library Cave
To understand Stein’s annotations, it is first important to be clear about how and when he acquired material from the cave, especially as there were at least four separate times of acquisition: we call these ‘acquisition events’.18 These are listed (in the form Acn, Ac for Acquisition, followed by number), dated—when known— and explained below. Ac1 is subdivided into four parts, important for understanding the annotations, as discussed later. While in a very few cases Stein’s published and unpublished records do not always tally exactly and are not always clear, they nevertheless give an almost complete and consistent account.19 Not included in this list is his initial brief visit to Mogao on 16 March 1907 when Wang was absent and Stein was shown one Chinese scroll by a young monk (1912: vol. 2, 20–31).
Ac1: Dunhuang Mogao, 22–30 May, 1907 [see diary and notebook MSS Stein 141]
Stein arrives 21 May 1907 ‘by 4pm’. (MSS Stein 204/355). On May 22, he takes photographs of the giant Buddha and other caves. Wang shows Jiang the entrance to the cave which he has bricked up and later that night gives Jiang a bundle previously removed from the cave to take to Stein’s tent ’for inspection.’ May 23: Wang has opened the entrance, shows Jiang and Stein the cave and agrees to show them some material. Stein and Wang move to the antechamber to Cave 16 which had a door to the corridor and papered-over windows to the outside so that they could not be seen by any visitors. Wang brings material to them which they examine during the day. The start of the entry in Stein’s notebook (MSS Stein 141/58) reads:
‘First explored huge MSS. Hoard. Convolute with Skr. Pothis emerged at 3P.M. Texts in confusion. Large MS 21” × 4”. C.A. [Central Asian] – Prototibn. [Prototibetan] MS. 14” × 2”. Brahmi compl. 32ff.
8 packets of Pothi foll.
1 packet of large sheets (3)
2 Tibetan rolls.
Painted cloth, 28” × 18”, perhaps intended for book wrapper. Chin. Inscriptn.’
Wang was bringing them ‘miscellaneous bundles’ which Stein surmised that he had left until last when replacing material in the cave after his initial discovery and emptying of the cave.20 As mentioned above, material in these bundles might also have come from different caves and/or been the remaining contents from bundles which Wang had previously unpacked for gifting.
After Stein and Wang had made their selections in Cave 16 during the day and left there ‘a big bundle of properly packed manuscripts and painted fabrics’ (1912: 179) they returned to their quarters—Stein to his tent pitched in front of the caves and Jiang to the camp he had made at the foot of the large Buddha in Cave 130.21 Their daily selections were then taken by Jiang to Stein’s tent under cover of darkness. The bundles contained many paintings and banners as well as material in languages other than Chinese or Tibetan (1921: 801ff). (Ac1a)
Wang then mostly emptied the cave and stacked the bundles in the corridor of Cave 16 (by nightfall 28 May) and these were counted and photographed by Stein (Figure 3). Stein and Jiang set aside paintings and other non-Chinese textual material from the bundles. Stein then examined the few bundles left in the cave and extracted more paintings and textiles on May 28 and 29 (1912: 193). On the morning of May 29 Stein and Jiang took their latest selections to Wang. Wang had meanwhile returned the bundles to the Library Cave ‘except 30–40 less well preserved rolls.’(MSS Stein 204/319). The material replaced by Wang in the cave seems to have included some previously selected by Stein as he writes ‘but I managed to extract from it [assuming here he means the Library Cave] about half of formerly selected bundles. First taken by Sieh [Jiang] to be added to 54 now laid out and marked.’(Ibid.) He then gives an inventory. In addition Wang also presented them with ‘several broken bundles, also large embroidered Buddha curtain, app. lifesize.’ The latter is the piece now commonly known as the Vulture Peak embroidery.22 ‘Transaction settled by 11.10 A.M. to mutual satisfaction’, ie payment agreed. At 10.30 pm, when it was dark, Jiang went with two others to retrieve the material agreed and they returned ‘by streambed’ to Stein’s tent at 1130pm with ‘two huge bag fulls. Brought in two more Taghars [bags] on second trip…. Bags left outside tent at 1215 am and carried to house later.’23 (MSS Stein 204/321)
Figure 3
The double exposed image showing some of the bundles of manuscripts in the corridor of Cave 16. Stein 1907. The British Library, Photo 392/59(2).37
On May 30 the MSS were placed in the storage hall and Wang visited to thank Stein for the payment. He then left for begging alms in the town. Stein visited and photographed some of the rock-cut temples, sorted and packed the manuscripts and paintings (note on June 3, 141/61) and also took photographs of some material (June 4, 141/61 and see British Library, Photo 392/27(565, 567, 580ff)).
On June 6 Wang has returned and Stein was allowed to make further selections from the miscellaneous bundles: ‘previously overlooked silk paintings eg C.lx.-C.lxvi.- also some others from C.i.’ … ‘Which I had left some fine silks in shyness the first day’ (MSS Stein 204/338). (Ac1b) and was also presented with ‘13 bundles of Chinese rolls containing mainly Chinese scrolls and Tibetan pothi.’24 (Ac1c) There was a further group of miscellaneous bundles below these which also contained many paintings and textiles, which Stein set aside and to which he gave site and bundle numbers to over the following day. Regular bundles were added by Wang to this material during the negotiations for purchase, but there was no opportunity to give these site-marks at the time.’ (Ac1d: probably about 40–50 bundles, see below). In total, Ac1a-d consisted of over 300 paintings/embroideries, a case of Tibetan and other non-Chinese material, and 90 bundles of Chinese/Tibetan manuscripts containing about 1200 items which altogether were packed in 16–17 cases.25
Ac2: Anxi, 1 October, 1907 (see diary)
Jiang returns to Dunhuang from Anxi and acquires more bundles from Wang which he takes to Anxi, variously 220 or 230 bundles, approx. 3000 text rolls … mostly Chinese sutras and Tibetan buddhist works .. single Sanskrit MS. on palm leaf.’ (MSS Stein 4/139).26 These were originally packed in huge bags and transferred to crates in Khotan in July 1908 (see below). They were almost certainly not given any numbers until Khotan (see Id2 below).
Ac3: Dunhuang Mogao, 2–7 April 1914
570 Chinese scrolls: five cases (MSS Stein 216/28, 7/4/14: AS to PA, 11/4/14, MSS Stein 11/49–50).‘each as large as a pony could conveniently carry.’27 (Stein 1928: 358) In the packing list for the 3rd expedition, he lists these as bundles, namely 53 bundles and detached rolls from 5 bundles, although this presumably includes the material from Ac4, at least 46 rolls.
Ac4: Various, 1913–16
Stein mentions purchasing manuscripts with Dunhuang as their purported provenance during his 3rd expedition from several sources. The first were offered to him by ‘an unknown Chinese well-wisher’ on his arrival at Dunhuang town on 1st April 1914, ‘Ching [rolls] 21’ for which he paid a ‘modest’ sum (not recorded in his accounts unless included in the payment on the following day). The same person returned the following day with ‘fresh bundle of c.25’ for which Stein paid 5 taels of silver.28 Stein also noted that many officials, including McCartney, had such manuscripts.29 It is not clear whether he was given any of these. It is possible that some of these were forgeries.
C: Stein’s Inventories of his Acquisitions
Stein had a system for inventorying his finds. When he visited a new archaeological site, most previously unrecorded, he would give it a site id. So, for example, the site of Niya, the kingdom of Cadota in the southern Tarim to the west of Dunhuang, was N. He would make a survey and plan of the site and assign an upper-case roman numeral to each discrete location at the site, such as a house or temple, so N.I., N.II., marked on his plan. If applicable, he then used a lower-case roman numeral to indicate the rooms or sections of the house/temple. Any finds discovered were numbered sequentially after this and annotated with this site id. This work was usually carried out every evening by Stein and he also prepared separate lists of the items. When the items were later sent to London, his assistants there could check off the annotated objects as they unpacked them against the lists. Any missing items would immediately be evident. The excavated items were then listed in the index to his expedition reports under this number.
Stein was generally meticulous in this work and curators and scholars have benefitted from it since. Given the amount of material and logistics in the field, there are surprising few cases when items are not annotated. But the rock-cut temples at Mogao, Dunhuang, was not an archaeological site. Although much neglected it was still active as Stein himself noted.30 And, as described above, Stein acquired the material from the library cave in rushed and clandestine circumstances. Unlike other material, therefore, there was no complete master list to use for checking when unpacking the Dunhuang manuscripts and paintings in London, nor were all annotated with a site id.31 Consequently, relatively few, about 600 out of several thousand, were listed in the index to Serindia. Nevertheless, Stein tried as far as possible to annotate and document what he could in the time available. In order to understand this documentation, we found it helpful to break down the systems used by Stein and others over the periods of acquisition and when and to which groups of material they were applied. For convenience, we label these identifications as Idn.
Id1: Ch.i.001, ie Ch. following by Roman number, lower case, and then Arabic number with two leading zeros. We also include under this category Id1a and Id1b, in both of which the final serial number is missing, eg Ch.ix.. The second part is sometimes given in upper case Roman numerals, see especially Tibetan manuscripts. In some cases, Ch is not present or C is used. Stein noticed this when visiting the British Museum in 1910 and clarified that all Mogao material should be prefixed ‘Ch.’ (MSS Stein 147/20). Listed in the index of Stein’s 1921 report. (See Figure 2a-d).
Id2: mazi number, usually in red, often also in Arabic numbers in blue pencil. Not in 1921 index (but see below). (See Figure 4).
Id3: Ch.1 ie. Ch. followed by number with no leading zeros. In 1921 index.
Id4: Ch. + Arabic numeral + Roman numeral + (Arabic numeral), eg. Ch.78.X.(1). Not in 1921 index.
Id5: Ch.001, ie Ch. with number with two leading zeros. In 1921 index. Chinese/Khotanese/Sanskrit
Id6: Ch.01, ie Ch. with one leading zero, used on 2nd and 3rd expedition. In 1921 index.
Id7: Chien.0001, ie Chien. followed by number with 3 leading zeros. Not in 1921 index.
Id8: other numbers written on the manuscripts in various formats.
Figure 2
Examples of Id1: the top three, are almost certainly in Stein’s hand and made in situ; the bottom two were made on unpacking in the British Museum or copied from Stein’s annotation when the item was conserved. From 1919,0101,0.125; 1919,0101,0.162; IOL San 389; IOL Khot 19; IOL Khot 2. The bottom two also show the addition of cataloguing numbers.
Figure 4
Detail of the verso of manuscript Or.8210/S.101 showing the mazi number in red and title in black added by Jiang and the corresponding Arabic number in blue pencil added on unpacking or later. The British Library.
A survey and documentation of the annotations on all the material in the Stein collection is not yet complete but we are confident that this list is comprehensive as it is based on an examination of several thousand manuscripts as well as the other documentation and working with the manuscripts and other material for many years. Ch. is the site id. assigned by Stein, short for Ch’ien-fo’-t’ung, the common romanisation at the time for the Chinese 千佛洞 ‘Thousand Buddha Caves’. Below we examine the identifications and try to identify when, where and by whom they were applied.
Id1 (eg. Ch.i.001): is the system which Stein applied on his second expedition to the manuscripts, paintings and textiles retrieved first from what he called the miscellaneous bundles, Ac1a, and which he had time to annotate on the ground, as was his usual practice. He originally used ‘C’ for Ch’ien-fu-t’ung, as seen in his diary and on some manuscripts, eg. C.i. - C.lvii. given in his diary (MSS Stein 294/308–17). It is probable that he also labelled items in Ac1b and Ac1c at this time (but see below), with the same system, but the system then fell apart owing to the amount of material and the very limited time available. This id. records the material by bundle, as noted by Takeuchi, the bundle number shown by the Roman number (2012: 210). The items in the bundle are then given a serial number, 001 etc. However, the serial number was added on unpacking at the British Museum.32 Items with id. in this format are listed in Serindia as ‘paintings, woodcut, textiles, and miscellaneous antiques’ (Stein 1921: 544) from Dunhuang, eg. the sutra wrapper mentioned above, MAS.858 (Ch.xlviii.001).
In some cases, the final part of the id. is not given: we only have the site and bundle number, the latter sometimes in lower case and sometimes upper case, eg. Ch.ix (Id.2a), Ch.IX (Id.2b). In others, the number is on later repair paper and in a different hand: but this almost certainly copied from Stein’s notation which was lost or covered over during the repair.
The last bundle number found recorded on Chinese, Khotanese and manuscripts other than Tibetan is cvi (106).33 Tibetan manuscripts contain ‘CCC’ but these notations are not in Stein’s hand and evidence suggests that these were added later by the cataloguer (la Vallée Poussin 1962: 109) and are not part of Stein’s original system. The Tibetan manuscripts show several such discrepancies, discussed below.
From Stein’s diary we can see the bundles he looked at each day: May 23rd: C.i. to C.ii.; 24th: C.iii.to C.xxi.; 25th: C.xxii. to C.xxx; 26th: C.xxxi to C.xli; 27th: C. xlii to C.xlvi; 28th: C.xlvii. to C.lvii (MSS Stein 294/308–17).Thus the bamboo sutra wrapper mentioned above, MAS.859, which has the id. Ch.xx.006, was one of the items selected on 24th May 1907 from bundle 20 (xx) while viewing material brought to him by Wang Yuanlu in the antechamber to cave 16. It was then the sixth item annotated from this bundle by Stein’s assistants when unpacked at the British Museum.
The miscellaneous or mixed bundles were brought out first because they were on top, placed there, Stein assumed, by Wang after he had emptied and then refilled the cave, because of ‘their very irregularity.’ This suited Stein as they contained many manuscripts, in both scroll and pothi format, with Brahmi script, generally to transcribe Sanskrit and Khotanese languages. They also contained many Tibetan pothi mixed up in varying sizes ‘in great confusion’ (1921: 816) as well as texts in Old Uyghur and Sogdian. As Stein notes, because of lack of time, ‘no classification of the different objects (manuscripts, paintings, decorated fabrics etc.) found in the same bundle could be attempted.’ (Stein 1921: 814 n.2).34 Nor, as Stein admits, was this system comprehensive: when numbering took place later many paintings and manuscripts were not part of any bundle and therefore simply given a serial number, preceded by two zeros (Id5 below, eg Ch.001)35
As noted above, Stein spent from 23–27 May examining these miscellaneous bundles, interested in anything dated and in languages and scripts other than Chinese. On 28 May, Wang Yuanlu removed the remaining regular bundles of Chinese and Tibetan manuscripts from the cave and placed them in the corridor of Cave 16 where they were photographed by Stein (Figure 3), although this negative was double-exposed.36 Stein also counted them, noting there were 1,050 Chinese bundles each containing about a dozen manuscripts, as well as eighty packets of Tibetan scrolls and pothi and eleven large Tibetan pothi (Stein 1921: 822–3).
There were more large mixed bundles originally beneath these in the cave which Stein examined on 28 and 29 May and assigned bundle numbers Ch.xlvii-lvii. He hypothesized that these had been placed at the bottom of the cave by Wang to level the floor: whether this was their original position in the cave is impossible to say. These largely contained paintings which he noted were very difficult to extract because the bundles had been crushed. He also selected more material from the earlier bundles at this time and presumably annotated them accordingly. The additional bundles in Ac1c and Ac1d were not annotated at this time.
Id2: refers to annotations made in red found on a proportion of the Chinese manuscripts which use a Chinese numbering notation often called Suzhou mazi, which was the traditional system used in accounting and mathematics in China (Figure 4). Many manuscripts also have this as an Arabic number written in blue or black pencil. Given the use of this Chinese system, these numbers were almost certainly written on the manuscripts by Jiang. The start of the sequence shows Chinese titles written in red ink, same as the mazi, and probably in Jiang’s hand. As Stein records, this work was done in Khotan in June-July 1908 at the end of the expedition and before the material was packed for dispatch to London (1921: 916 and Figure 5).38 The Arabic numbers were presumably added in London to help the non-mazi reading assistants. It is possible that some were added when the material was unpacked on the visit of Paul Pelliot in 1910.39
Figure 5
Jiang Xiaowan annotating the Chinese material near the end of the second expedition. Khotan, 31 July 1908. The British Library, Photo 392/26(831).
Jiang was most probably working on Chinese material in Ac1c, Ac1d and Ac2 which had not been annotated by Stein. Stein’s packing list (for case 72) states that Jiang reached number 1318 (MSS Stein 37/119).40 Stein notes also that Jiang prepared an index slip list of these 1318 manuscripts, which presumably included the number and title and which was packed in box 91, but this has not been located. No bundle information is included with these (but see below).
Further support for these mazi numbers being added by Jiang while in Khotan is given by lists of dated manuscripts he came across while doing this numbering and which Stein recorded in his notebooks (eg. MSS. Stein 272/59, Figure 6). So, for example, on 27 June 1908 in Khotan Stein noted the manuscript 778 which had the date 743. This corresponds to Or.8210/S.316 which, indeed, has 778 written in mazi in red ink and in roman numbers in blue crayon.41
Figure 6
Stein’s notes from 27 June 1908 in Khotan listing dated manuscripts identified by Jiang (MSS Stein 272/59).
Id3 (eg. Ch.1): Found in the text and index to Serindia but not generally on items in this form (but see below on Tibetan) and used for eighteen manuscripts selected with the help of Paul Pelliot and Lionel Giles for reproduction in Plates CLXVI to CLXIX in Serindia because of their ‘exactly fixed dates, their palaeographical features or contents, or for other reasons may claim some special interest.’ (1921: 917–8). The number conforms to the mazi number [Id2]. So, for example Or.8210/S.2199, a nun’s will dated to 865, has the mazi number 1283 on the manuscript itself. It is referred to as Ch.1283 in Serindia (1921: 918 and Plate CLXVIII): similarly, Or.8210/S.367 [Ch.917], Or.8210/S.1524 (Ch.1181] and Or.8210/S.2724 (Ch.401). In some cases, we have not found a corresponding mazi number on the manuscript but this is probably a case of it having been inscribed on a section of the manuscript which has since been lost, e.g. a fragmentary section at the beginning, or on original backing paper, since removed.42 So we can hypothesize with some confidence that Id2 and Id3 give the same information.
Using the mazi number with the Ch. prefix gave them a site identification before systematic numbering was carried out. Because the numbers after Ch. correspond to existing mazi numbers, they are not consecutive. They are all Chinese.
Some Tibetan manuscripts are also inscribed with an id. which starts with this form, eg IOL Tib J 37 has Ch.2.D. And IOL Tib J 42, Ch.2.fr 1. These are not in Stein’s handwriting nor part of this sequence.
Id4 (eg. Ch.78.ix.001): This id. incorporates the site id with the number of the case in which the material was packed in Khotan from June 9 to July 18, 1908, to send to London: Stein prepared a packing list for each numbered crate (MSS Stein 37/117–20, Figure 7).43 So, for example, Ch.78.ix.001, is the first item from bundle 9 found in packing case 78. This only seems to have been used for the manuscripts and not for paintings, most of which had an id1 (the paintings were packed in cases 41 to 45 and the manuscripts in cases 67 to 91. Cases 67 to 71 contained Chinese manuscripts numbered by Jiang in Khotan, ie Id2, and therefore did not receive an Id4.44 Case 72 contained manuscripts with Id2, but also two additional bundles which were presumably given an Id4., eg Or.8210/S.692 which has the Id4, Ch.72.VIII.22.
Figure 7
Part of the list of packing cases from 2nd expedition prepared by Stein in Khotan and with annotations made on unpacking at the British Museum in 1909. The Bodleian Library. MSS Stein 37/120.
These numbers were added to the manuscripts on unpacking in the British Museum by Stein’s assistants there— as we have seen from Stein’s comments above, there was no time to carry out this annotation in Khotan before they were dispatched. The cases started to be unpacked in 1909 (MSS Stein 37/116) but the numbering continued at least into 1914.45
The ids. on the manuscripts confirm to this system up to case 87. But after this there is some confusion. There is one manuscript on the database marked as from case 88, namely IOL Tib J 753, with id. Ch.88.VI. The fact that the bundle number is VI. Suggests that there should be others with bundles number I-V. It is possible that these were Chinese manuscripts and these ids have not yet been recorded. But this is strange as, according to Stein’s packing case lists, cases 73 to 91 contained Tibetan material. We might expect a a few Chinese manuscripts mixed among them, but not many. Similarly, case 89 has only two manuscripts, both Chinese: Or.8210/S.2217 with Ch.89.V.5 and Or.8210/S.7111 with Ch.89.XII.13/19. Again, the bundle numbers suggest many more manuscripts were annotated with this case number. There are no manuscripts with case number 90, and four —none of which are Tibetan —with case number 91 and bundles numbers I, II, and VII.
So where are the missing Tibetan manuscripts? There are 90 Tibetan manuscripts with case 9. We know from Stein’s lists that case 9 did not contain manuscripts, so it is a reasonable assumption that these might belong to case 90 or 91. Looking at the id written on these, it becomes more confusing as several of them have clearly been changed from the original annotation or are in different handwriting from many others (see Figure 8a–d). So IOL Tib J 588, for example, was probably Ch.IX. There are also two manuscripts with Ch.8, one Chinese and one Tibetan, and fifteen Tibetan ones with Ch.2. The last are in a different format, so not followed by a bundles number in Roman numerical, but by a capital letter or fragment number: Ch.2.E. or Ch.2.fr.1.
Figure 8
Paper scroll wrapper from Dunhuang, Or.8210/S.11049 with annotations by Jiang Xiaowen. The British Library.
As Stein notes of the Tibetan material, ‘owing to a variety of causes, including their great mass which is second only to that of the Chinese, no sufficiently extensive examination has as yet been possible.’ (1912: 218) It is not certain that all of them were numbered immediately on unpacking, possibly because of the fact that, unlike the Chinese scrolls, which each received a distinct number even if they were, strictly speaking, only sections of a longer text which consisted of several scrolls, it was more difficult for the assistants in the British Museum to distinguish distinct texts or sections of text among the mixed up individual pothi leaves. The assistants did not read Chinese or Tibetan. It is therefore possible that they were set aside at this initial unpacking until they could be sorted by relevant scholars and given other ids later and that those from cases 90 and 91 were mainly conflated. Some Tibetan manuscripts were sent to Dr F. W. Thomas (1857–1956), a Tibetologist and Librarian of the India Office, in November 1909.46 Manuscripts started to be sent to Louis de la Vallée Poussin (1869–1938) from October 1910 after Stein and he had met in Paris (MSS Stein 102/153). Slips, made in London, were also sent for identification.
Caroline Ridding (1962–1941) looked at some of the manuscripts in this early period and then worked with la Vallée Poussin on cataloguing them between 1914–18 when he was living in Cambridge (although la Vallée Poussin was named as sole author of the subsequent catalogue 1962).47 It is probable that the manuscripts were sent to Cambridge for this.48 Thomas was to inventory them for the India Office Library, to which they were all assigned and most of the volumes contain index cards signed by Thomas dating to the 1920s which give the LVP catalogue number and id.49 Looking at a random sample shows that the ids recorded in la Vallée Poussin do not always appear on the manuscript; or are partial and sometimes on different pothi. They are also in different handwritings, not Stein’s.50 All of these anomalies support the hypothesis that they were not systematically annotated on the initial unpacking. It might be possible to make more sense of these ids with further study, including comparison of handwriting, but this is the subject for another paper.
Id5 (eg. Ch.001): used for ‘for 2nd expedition ‘objects which were recovered amongst manuscript bundles etc., at the British Museum’ but not obviously part of any bundle (1921: 836, n. 13). Stein goes on to say, they ‘are marked with consecutive numbers, preceded by 00, after the general ‘site-mark’ Ch. (e.g. Ch.0017, Ch.00452 etc.).51 Items with this id are listed before those with Id1 in the Serindia list ‘paintings, woodcut, textiles, and miscellaneous antiques’ from Dunhuang. This id. was only inscribed on the objects at the British Museum. The published list continues to Ch.00522. The database of Stein items in London collections includes 691 items with this Id. but many are multiple fragments with same id. no.
Stein noted some of the non-Chinese material with these ids when in the British Museum in August 1917 (MS Stein 272/132–133) His comment that these were ‘not obviously part of any bundle’ is not strictly accurate. There are several manuscripts which contain an id4 and an id5, such as Or.8210/S.2404 and IOL Khot S 14. In both cases the ids are written next to each other in different hands. In the latter case, id4 is scored out.
Id6 (eg. Ch.011): This was used by Stein on both his second and third expeditions. In the second expedition it was used for eleven Tibetan manuscripts items, all pothi except Ch.011 which was a collection of scrolls. They appear under this id. in Appendix I of Serindia (1470–1) prepared by Thomas of the India Office, from the catalogue prepared by la Vallée Poussin. The manuscripts are shown on Plates CLXXIII and CLXXIV. They are now catalogued in the IOL Tib J sequence. The Appendix notes that Ch.011 was part of ‘bundle 86.1–4: P. 99’ suggesting this was originally a separate sequence as is suggested by an outlier, Ch.0180, in the index which refers to a Tibetan manuscript, perhaps suggesting that the Tibetan manuscripts had been assigned Ch.012 to Ch.0180, if not beyond.
The same id. was then used by Stein on his third expedition for a small group of clay plaques which were given to him by Wang in April 1914, Ch.01–034 (1928: 359, 361–2).52 Most of these are now in the New Delhi collection, eg 99/7/19 is Ch.024.53
Id7 (eg. Chien.0599): material from Stein’s 3rd expedition (Ac3) in which he used the alternative site abbreviation, ‘Chien.’ to distinguish this material from that obtained on previous expedition. This is also corroborated by the position of most of this material at the end of the registration sequence, assigned at the British Museum, of the form Or.8210/S.nn Although including the prefix, S.8210, which generally refers to material from the 2nd expedition, the manuscripts marked with ‘Chien,’ are primarily found at the end of this sequence, so from Or.8210/S.6350 onwards. This was noted by Fujieda Akira (1911–1998) who believed that many of these were forgeries.54 It is also clear when looking at the manuscripts in storage. The manuscripts get smaller and more various, including booklets etc in various languages. Then, around Or.8210/S.6350, we see complete scrolls again in Chinese, some very long. This concurs with Stein’s observation when negotiating for this material in Dunhuang in 1914.55 In Innermost Asia Stein shows photographs of eight of the scrolls, all labelled ‘Chien.nnnn’ (1928: Plates LXV, LXVI).
The highest Chien. number we recorded during our initial survey was Chien.0599. This is slightly more than Stein’s count of 570 rolls but it is probable that he includes that c.46 rolls purchased from his ‘well-wisher’ [Ac4] with this material.56 There is probably little information to be had from these about the original order of the scrolls unless we assume that the Chien numbers were given sequentially as each bundle was opened so that we can assume, for example, that Chien 1–10 might belong to the same bundle. But it is still then uncertain how many were in each bundle and so when each bundles starts and ends, so the system soon falls down. They are therefore not discussed further here.57
Id8: some items bear different or additional numbering. For example, Or.8210/S.529 has 912 in mazi and Arabic numbers. but also 882 written in black pencil. Sometimes these refer to the catalogue number as given by Giles for the Chinese manuscripts. In other cases, their purpose is not clear, but it is most probable that these were inscribed on the material when it was in the British Museum and being sorted and catalogued. They are not discussed further here.
Some manuscripts contain ids from more than one system. For example, Or.8210/S.6263, has the numbers Ch.00187 and 80.XI, ie id1 and id5. Given the length of time taken with numbering and the number of different people working on this material this is not surprising. Some cases are discussed below.
D. Testing the documentation
To some degree, we can test the validity of the above analysis from the types of material found with the different id numbers. We would expect material with id1&4 to contain manuscripts in all languages along with paintings and banner material, id2-3 to be almost, if not exclusively, Chinese, whereas those with id5 to be mainly smaller, fragmentary pieces. If this model holds up, we can potentially use the ids to reconstruct some of the bundles which Stein acquired. Stein certainly believed this to be the case. On the examination of material after it was sent to the British Museum, he notes: ‘the way in which banners belonging to particular groups have turned up from bundles bearing proximate numbers proves the utility of the care taken in preserving the original ‘site-marks’, given by me to the bundles successively brought to light. It also shows that the original contents of the bundles are not likely to have been mixed up altogether by Wang…’ (1921: 852, n.33).
First, to continue Stein’s reasoning and to test further the above interpretation and accuracy of the documentation, we look below at samples from each id.
Id1
This gives information on the contents of the miscellaneous bundles which, Stein notes, were on top of the cave placed there, Stein assumed, by Wang after he has emptied and then refilled the cave, because of ‘their very irregularity.’ He also notes that these might have contained objects brought from other caves, as discussed above. Wang Yuanlu brought them out first for Stein. This suited Stein as they contained many manuscripts, in both scroll and pothi format, with Brahmi script, generally to transcribe Sanskrit and Khotanese languages. They also contained many Tibetan pothi leaves mixed up in varying sizes as well as texts in Old Turkic and Sogdian.
Despite the uncertainty over whether they were in this form when originally placed in the cave, it would still be useful to reconstruct these to see if there is any pattern. Since most of this material consists of paintings, non-Chinese documents and other objects, such as the sutra wrapper, id1 has been comprehensively recorded in the various catalogues. Below is what we know of Ch.i. from the IDP database, the catalogues and the index to Serindia — It might be useful if a reconstruction of each bundle was attempted in this way.
Miscellaneous bundle 1, Id1, namely Ch.i.
This bundle was brought to Stein on 23 May, and he records that he selected the following 14 items:
‘6 silk gauze paintings, small;
2 painted cotton flags, small
3 large silk paintings;
2 paintings on paper;
1 large painting on cotton.’ (MSS Stein 204/306–8)
He further notes that on 6 June he made further selections
‘some others from C.i. …which I had left some fine silks in shyness the first day.’ (MSS Stein 204/338–339)
If we look at Serindia, Ch.i.001 to 023 are recorded, all paintings except Ch.i.021. Ch.i.021 consists of four items, a-d, namely Khotanese and Sanskrit manuscripts. These can be readily identified with manuscripts in the BL, namely IOL Khot S.19-S.21, S.47 and IOL San 389. Four of the paintings are on paper (009, 017, 018 and 023); three noted by Stein as on linen (014, 015, 016); the remainder, 15 items, are banners and paintings on silk. This loosely tallies with Stein’s list above. But there are some issues. Interestingly, the location of Ch.i.001 cannot be confirmed: it does not tally with any of the paintings listed in the BM or NMI collections. It was described in Serindia, published in 1921, but the manuscript for this was finished before the paintings were acquisitioned into the BM collections in 1919, Stein working on it during his extended stay in England from 1916–17. The painting is not listed in the report produced by Raphael Petrucci by 1915 on a division of the collections (Wang 2004: CE32/23/62/2&3).58 To add to the confusion, there are other items with this id: namely Tibetan manuscripts (IOL Tib J 310–1034 and IOL Tib J 363). We discuss the issue of variant Tibetan ids below. The seemingly mislaid painting is more problematic and will be explored further in a future article.59 However, broadly speaking the contents of bundle I. confirm to Stein’s description as given on the ground. It might be worth looking at them in more detail to see if there is anything linking their contents—as Stein suggested for the banners from one bundle, quoted above—but we leave that to others.
Id2 and id3
Although no specific bundle number is recorded on these items, it is reasonable to assume that Jiang worked on them by unwrapping and numbering a bundle at a time. This is supported by the annotation of two of the few surviving wrappers with a sequence in mazi. The first is a hemp wrapper, Or.8210/S.11464, and Jiang’s inscription reads ‘Tibetan sutras (番經) from 1081 to 1098’. This is interesting as it suggests that Jiang might have annotated some non-Chinese material. Currently, we have no records of mazi in this sequence on the database, but they have not been checked so this does not mean it does not appear on some. The second, however, is more helpful. It is a wrapper made of layers of paper, the outer layer having a blue-dyed border (Or.8210/S.11049, Figure 8). According to Jiang’s annotation, it contains Id2, 605 to 622. Fourteen manuscripts with mazi numbers in this sequence can be identified in the database: see Table 1.
Table 1
Manuscript scrolls so far identified as contained in the paper wrapper pictured in Figure 8.
| MAZI | BL MSS. NO. | TEXT (GILES 1957 NO.) | DATING |
|---|---|---|---|
| 605 | Or.8210/S.238 | Daoist text: Jin zhen yu guang ba jing fei jing 金眞玉光八景飛經 (6834) | Empress Wu characters in colophon |
| 606 | Or.8210/S.244 | Saddharmapuṇḍarı̄kasūtra, ch. 6, sections 23 (end only), 24 (3024) | |
| 606 | Or.8210/S.245 | Saddharmapuṇḍarı̄kasūtra, ch. 2, sections 3, 4 (2429) | |
| 606 | Or.8210/S.246 | Saddharmapuṇḍarı̄kasūtra, ch. 5, section 17 (2891) | |
| 606 | Or.8210/S.247 | Saddharmapuṇḍarı̄kasūtra, ch. 6, sections 18, 19 (2910) | |
| 606 | Or.8210/S.248 | Saddharmapuṇḍarı̄kasūtra, ch. 9, sections 22, 23 (3008) | |
| 607 | Or.8210/S.243 | Mahāprajñāpāramitāsūtra, pt. 1, ch. 31, pin 7 (21) (41) | |
| 609 | Or.8210/S.241 | Vimalakı̄rtinirdeśasūtra, ch. 2, sections 8 (end only), 9 (3442) | |
| 612 | Or.8210/S.235 | Buddhāvataṃsakasūtra, ch. 48 [=ch. 58 (mid.)] (1645) | |
| 616 | Or.8210/S.229 | Sarvadurgatipariśodhana Uṣṇı̄ṣa Vijaya Dhāraṇı̄ Sūtra (N.348) (3828) | |
| 618 | Or.8210/S.236 | Li chan wen, 禮懺文 (6468) | 981? |
| 619 | Or.8210/S.232 | Guan shi yin bu kong juan suo wang xin shen zhou gong de fa men ming bu kong cheng jiu wang fa 觀世音不空羂索王心神呪攻德法門名不空成就王 (5161) | |
| 621 | Or.8210/S.228 | Saddharmapuṇḍarı̄kasūtra*, ch. 2] section 3 (2375) | |
| 621 | Or.8210/S.233 | Suvarṇaprabhāsottamasūtrendrarājasūtra*, section 16, 17 (2110) |
This suggests that we would therefore be able to retrieve some bundle information as contiguous numbers would, in many cases, belong to the same bundle. It is difficult to test this further until all the mazi numbers are recorded, but we can say that this sequence may provide some useful information.
Id4
This is probably the most useful id to explore as it describes material kept in what Stein called the ‘regular bundles’ in the body of the Library Cave, those whose contents are most unlikely to have been substantially disturbed or rearranged by Wang and thus most likely to be as originally placed there. At present, we can only do an initial test as the database does not yet contain complete data on this id. But, looking at Ch.76, i.e. material in case 76, we find 202 records on the database with records for 11 bundles, i.e. c.18 items per bundle. Stein described this case as containing 12 bundles of Chinese manuscripts, unsorted. It is not clear if the case originally contained 12 bundles and one of these was accidentally mixed up with another on unpacking; if it contained material which has not yet been recorded on the database; or whether Stein’s original packing list is mistaken. We look at one bundle here (Bundle A and see Table 2 below) and discuss this and two other bundles (B and C) in the next section.
Table 2
Bundle A: Contents of Bundle 9 in case 79 (Ch.79.IX).
| STEIN BUNDLE | BL REG. NO. | TITLE (GILES 1957 NO.) | HEIGHT/cm | LENGTH/cm | COPIER |
|---|---|---|---|---|---|
| Ch.79.IX.1 | Or.8210/S.1875 | Sukhāvatīvyūhasūtra (4911) | 30.8 | 528.22 | Lu Rixing |
| Ch.79.IX.2 | Or.8210/S.1834 | Sukhāvatīvyūhasūtra (4889) | 31 | 497.74 | Fan Zisheng |
| Ch.79.IX.3 | Or.8210/S.2019 | Sukhāvatīvyūhasūtra (4930) | 31 | 167.74 | |
| Ch.79.IX.4 | Or.8210/S.1838 | Sukhāvatīvyūhasūtra (4892) | 31 | 354.34 | Lu Rixing |
| Ch.79.IX.6 | Or.8210/S.2012 | Sukhāvatīvyūhasūtra (4923) | 31 | 815.34 | |
| Ch.79.IX.7 | Or.8210/S.1862 | Sukhāvatīvyūhasūtra (4901) | 31.2 | 172.52 | Song Sheng |
| Ch.79.IX.8 | Or.8210/S.1982 | Sukhāvatīvyūhasūtra (5092) | 30.5 | 205.74 | Wang Han |
| Ch.79.IX.9 | Or.8210/S.1990 | Sukhāvatīvyūhasūtra (4916) | 31 | 678.18 | Zheng Jing |
| Ch.79.IX.10 | Or.8210/S.1870 | Sukhāvatīvyūhasūtra (4906) | 31.2 | 167.64 | Piliang |
| Ch.79.IX.11 | Or.8210/S.2013 | Sukhāvatīvyūhasūtra (4924) | 31.5 | 213.36 | Song Sheng |
| Ch.79.IX.12 | Or.8210/S.1843 | Sukhāvatīvyūhasūtra (4898) | 31.2 | 497.74 | |
| Ch.79.IX.13 | Or.8210/S.1836 | Sukhāvatīvyūhasūtra (4890) | 31.6 | 599.24 | |
| Ch.79.IX.14 | Or.8210/S.1866 | Sukhāvatīvyūhasūtra (4902) | 31 | 213.36 | Song Sheng |
| Ch.79.IX.16 | Or.8210/S.1842 | Sukhāvatīvyūhasūtra (4897) | 31.4 | 213.36 | Ma Feng |
| Ch.79.IX.17 | Or.8210/S.1840 | Sukhāvatīvyūhasūtra (4895) | 31.2 | 213.36 | Ma Feng |
| Ch.79.IX.18 | Or.8210/S.2014 | Sukhāvatīvyūhasūtra(4925) | 31 | 441.96 | |
| Ch.79.IX.19 | Or.8210/S.1872 | Sukhāvatīvyūhasūtra (4908) | 31 | 162.46 | Lu Rixing |
| Ch.79.IX.20 | Or.8210/S.1837 | Sukhāvatīvyūhasūtra (4891) | 31.8 | 518.16 | Yao Liang |
| Ch.79.IX.21 | Or.8210/S.1868 | Sukhāvatīvyūhasūtra (4904) | 31.3 | 213.36 | Ma Feng |
| Ch.79.IX.22 | Or.8210/S.2015 | Sukhāvatīvyūhasūtra (4926) | 32 | 167.64 | Zhang Juanzi |
| Ch.79.IX.23 | Or.8210/S.1861 | Sukhāvatīvyūhasūtra (4900) | 30.2 | 182.88 | |
| Ch.79.IX.24 | Or.8210/S.1873 | Sukhāvatīvyūhasūtra (4909) | 31.5 | 175.26 | Lu Rixing |
| Ch.79.IX.25 | Or.8210/S.2018 | Sukhāvatīvyūhasūtra (4929) | 31 | 198.12 | |
| Ch.79.IX.26 | Or.8210/S.1869 | Sukhāvatīvyūhasūtra (4905) | 32.2 | 335.28 | Zhang Juanzi |
| Ch.79.IX.27 | Or.8210/S.2017 | Sukhāvatīvyūhasūtra (4928) | 30.5 | 162.46 | |
| Ch.79.IX.30 | Or.8210/S.1994 | Sukhāvatīvyūhasūtra (4920) | 31 | 731.52 | |
| Ch.79.IX.31 | Or.8210/S.1871 | Sukhāvatīvyūhasūtra (4907) | 30.5 | 335.28 | Lu Bao |
| Ch.79.IX.32 | Or.8210/S.1867 | Sukhāvatīvyūhasūtra (4903) | 31.3 | 335.28 | |
| Ch.79.IX.33 | Or.8210/S.1874 | Sukhāvatīvyūhasūtra (4910) | 30.8 | 213.36 | Song Sheng |
| Ch.79.IX.34 | Or.8210/S.1995 | Sukhāvatīvyūhasūtra (4921) | 30.5 | 213.36 | Song Sheng |
| Ch.79.IX.35 | Or.8210/S.1839 | Sukhāvatīvyūhasūtra (4993) | 31.5 | 213.36 | Song Sheng |
| Ch.79.IX.36 | Or.8210/S.1841 | Sukhāvatīvyūhasūtra (4896) | 31.2 | 198.12 | Song Sheng |
| Ch.79.IX.37 | Or.8210/S.1844 | Sukhāvatīvyūhasūtra (4899) | 31.1 | 167.64 | Fan Zisheng |
The highest serial number recorded on a scroll in Bundle A is 37 and the database search brings up 33 items; we can reasonably assume that the sequence is near complete. In addition, all the scrolls are also mostly complete — there are no fragments. They all contain the Chinese translation, Wuliangshoujing [佛說無量壽經], of a Buddhist sutra (Sukhāvatīvyūhasūtra, The Infinite Life Sutra), some containing up to four copies.60 All are on medium quality buff paper, c. 31–2cm high.61 Many have a copyist’s name at the end. These names are all found on several other manuscripts, all copies of this sutra. This would seem to support our conclusion and other documentation, that at least some of the bundles when acquired by Stein had retained mainly the same contents—even if Wang had removed some—in which they were placed in the cave. It would also support Rong Xinjiang’s conclusion that ‘these “regular library bundles” were in fact Buddhist sutras stored in traditional bundles.’ However, not all bundles are so consistent, as shown by the two bundles, B and C discussed below.
Id5
Searching the IDP database for this id brings up 691 items, most of which are fragmentary pieces of textiles, now in the V&A, or small paintings, now in the BM, which supports the narrative. But the sequence also contains manuscripts in Khotanese, Sogdian, Tibetan and Turkic. The Khotanese texts are almost all single pothi while the Tibetan texts consist of one to three pothi. All the Tibetan manuscripts contain at least one other number written in a different hand in pencil, so IOL Tib. J 667 is a single pothi from a Tibetan text. It also contains the annotations: ‘Ch.0017’ 641 B’ and ‘comp. 466’, the first in ink and the others in pencil, in different hands. ‘IOL Tib J 689 similarly has ‘Ch.0021’ and ‘670’; IOL Tib J 550, has ch.0033 and 527; IOL Tib J 75 has Ch.001 and 72. Although the second number is not the same as the eventual catalogue number, they are all close and so this could represent, we would suggest, an early ordering of the manuscripts for cataloguing prepared by la Vallée Poussin and Ridding between 1914–18. The work was only completed much later by F.W. Thomas and after both their deaths. The Khotanese manuscripts were not fully catalogued until much later, hence no sign of their early ordering in this way.
Several items also contain an id4: Or.82128/80(B), a Sogdian text which is annotated with ‘Ch.0093.B’ in red ink, which has also been used to score out the id4, namely Ch.87.XIII.001[?]’; Or.8212/81, Ch.83.IX and Ch.00349; and Or.8212/82 (Chinese and Sogdian), with Ch.82.XVI. and Ch.00335. These are identified in a list of Sogdian manuscripts by their id5 in Serindia (924), along with lists of Turkic Texts, also identified by their id1 or id5 (924–5). It seems as if these two ids— the first inscribed by Stein on the scene and the latter on their unpacking in the BM—were used in place of id4, despite thereby losing bundle information.
Although it was determined quite early that, in the division of material between the BM and India, the Chinese manuscripts would go to the former and the Tibetan ones to the latter, there was discussion about the other material until 1914.62 When Stein sent the proofs for Serindia, items such as Ch.0013, a Uyghur text, and Ch.0015, a Manichaean scroll, were assigned to India even though this later changed. The BM non-Chinese material was eventually acquisitioned with material from Stein’s third expedition (1916–18), given the Or.8212 prefix and the Keeper, Lionel Barnett, made a handlist (Barnett nd). He did not record the other numbers nor distinguish between 2nd and 3rd expedition material.
4. Restoring the Bundles
Stein noted four types of bundles: B1, miscellaneous bundles; B2, relics of worship; B3, library waste; and B4, regular bundles. It is possible to reconstruct at least some of these from his documentation. What he termed the ‘miscellaneous bundles’ (B1) were collections of material brought together by Wang Yuanlu, possibly some from other caves, and possibly some from bundles he had previously unwrapped. Further reconstruction of some of these bundles might be interesting to explore to see whether they can be linked to a specific shrine or period, although there simply might not be sufficient evidence.
The contents of B2, which he described as ‘relics of worship swept up from different shrines and put aside on account of religious scruple’ (1921: 820), do not have a specific bundle number and it would therefore be difficult to reconstruct them in most cases.
Stein further hypothesised that some of the larger bundles contained ‘library waste’ such as Ch.xxxv and xxxvii, which contained ‘torn ends of Sutra rolls stiffened with thin sticks of wood; of wooden rollers once used in manuscript rolls; silk tapes; cloth wrappers and similar library “waste”’ (1921: 812). Much of the fragmentary material found at the British Library in the pressmarks Or.6981 onwards fits this description, although some of this sequence also comes from paper fragments removed from other objects. It is unlikely that we can reconstruct these.
Turning to his final classification, B4, the regular bundles, were many of these, as Rong has argued, ‘Buddhist sutras stored in traditional bundles.’? (2013: 113) Rong based this conclusion partly on a photograph taken by Stein in 1907 and reproduced below, which shows a pile of bundles still in their hemp covers. An inscription on one of the bundles is clearly visible. It reads: ‘Mohe bore 摩訶般渃—海.’ Rong writes: ‘This is precisely the sutra name and bundle number according to the catalogue of canonical texts in the Kaiyuan lu 開元錄 catalogue. Indeed, this bundle must contain Bundle 2 of the forty-juan version of the Mohe bore jing 摩訶般渃 which appears under the character Hai 海.’ (2013: 112)63
Bundle A, examined below (Table 2), shows that indeed some bundles seem to have been comprised of Buddhist sutras (see Table 2 below), although further work would be needed to see if they were also catalogued according to Kaiyuanlu. Now that we can be reasonably confident that Stein’s Id4 provides information about the original contents of the regular bundles and, based on Rong’s argument and the example in Table 2, that some of them are collections of Buddhist scrolls, here we look at two more bundles for which we have reasonably complete data to see if they also fit this description.
The first is that comprising manuscripts with the id4 prefix, Ch.84.III, ie bundle 3 in case 84, hereafter referred to as Bundle B). The highest number recorded with this bundle id. is 14 and 14 items are recorded in the BL collection. They are all in Chinese and in scroll format and all but two are complete. However, as can be seen from Table 3, they do not confirm to any systematic Buddhist cataloguing system, comprising a range of Buddhist texts including both apocrypha and sutras from across the canon.64 Two scrolls are dated to the 7th century. Or.8210/S.1456 has a tabulated colophon, suggesting it was made in the capital of China, Chang’an. It is dated, to 29 June 676, and records that the editor, ‘Fajie (法界) is from Huadu Monastery’ (化度寺). Huadu is a known Sanjie (Three Stages) monastery in Chang’an and thus it is probable that the manuscript came from there. The Three Stages school developed in the sixth century and prospered during the seventh and eighth centuries in central China. Among the texts studied by its followers were some which are usually termed ‘apocryphal’ although, as Charles Muller (1998) has discussed, the term covers a variety of texts some of which were later accepted into the Chinese canon. The height of the paper, c.26–28 cm, suggests an early dating (Fujieda 2002: 104). But one has a club circular on the verso dating to 886, suggesting later reuse of this Buddhist manuscript by the lay community.65 The question then arises as to why this club circular ended up in a bundle with earlier manuscripts. Does this suggest that they were only bundled together after this date?
Table 3
Bundle B: Bundle 3 from case 84 (Ch.84.III).
| BUNDLE | BL MSS. NO. | TITLE (GILES 1957 NO.) | HEIGHT/cm | WIDTH/cm | DATE |
|---|---|---|---|---|---|
| Ch.84.III.1 | Or.8210/S.2686 | Foshuoxianghaojing 佛說相好經 (5344) | 198.12 | ||
| Ch.84.III.2 | Or.8210/S.1448 | Mahāprajñāpāramitāsūtra (71) | 26.5 | 883.92 | |
| Ch.84.III.3/4 | Or.8210/S.1450 | Buddhanama sutras (4684) | 27.8 | 266.7 | |
| Ch.84.III.5. | Or.8210/S.1445 | Pañcaviṃśatisāhasrikāprajñapāramitasūtra (783) | 26.5 | 441.96 | |
| Ch.84.III.6. | Or.8210/S.1444 | Mahāvaipulyamahāsamnipātasūtra (1590) | 25.4 | 701.04 | 7th c. |
| Ch.84.III.7 | Or.8210/S.2692 | Foshuofawangjing 佛說法王經 (5311) | 822.96 | ||
| Ch.84.III.8 | Or.8210/S.1456 | Saddharmapuṇḍarı̄kasūtra (2818) | 26 | 952.5 | 676 |
| Ch.84.III.9. | Or.8210/S.1446 | Buddhabhasitamahabhisekarddhidharanisutra (3550) | 25.9 | 533.4 | |
| Ch.84.III.10 | Or.8210/S.1453 | Vajracchedikāprajñāpāramitāsūtra (1118) | 27.2 | 499 | |
| Ch.84.III.11 | Or.8210/S.1447 | Caturvarga-bhikṣunı̄-prātimokṣa (4197) | 25.6 | 304.8 | |
| Ch.84.III.12 | Or.8210/S.1457 | Suvarṇaprabhāsottamarajasūtra (2059) | 26 | 426.72 | |
| Ch.84.III.13 | Or.8210/S.1458 | Visesacintabrahmapariprcchasūtra (3683) | 26.1 | 952.5 | |
| Ch.84.III.14 | Or.8210/S.1449 | Prajñāpāramitāhṛdayasūtra (1417) | 27.9 | 45.72 | |
There are many avenues to pursue here, including a more detailed look at the paper and wooden rollers—which might also help with dating—and at the mix of texts. Further research might suggest why this group of manuscripts were kept together.
Bundle 4 from case 80 (Bundle C) is another another bundle for which we have almost complete data, listed in Table 4, but which is much more heterogenous. It contains a mixture of Tibetan and Chinese material in different formats, comprising Buddhist texts, including Tantric invocations, and divination texts. The Tibetan texts and the concertina form of several of the manuscripts, point to a tenth century date, although one of the other texts contains characters mainly used during the reign of Empress Wu (r. 695–701), but not much thereafter. This possibly suggests a much earlier date for this manuscript—or it is a copy of an earlier manuscript by someone unaware of the changed conventions of regimes in China.
Table 4
Bundle C: Bundle 4 from case 80 (Ch.80.IV).
| BUNDLE | BL MSS NO. | TITLE (LVP/GILES/ENOKI CAT. NO.) | LANG. | FORMAT | HEIGHT/cm | WIDTH/cm |
|---|---|---|---|---|---|---|
| Ch.80.IV. | IOL Tib J 1589 | Aparimitāyurjñāna Sūtra | Tib. | scroll | ||
| Ch.80.IV. | IOL Tib J 384 | Maṇḍalakramavidhi and Amoghapāśamaṇḍala (LVP 384) | Tib. | concertina | 9 | 29 |
| Ch.80.IV. | IOL Tib J 716 | Tantric invocations, Vajrayāna (LVP 716) | Tib. | scroll | 25 | 158 |
| Ch.80.IV. | IOL Tib J 741.1 | [Not clear that exists] | Tib./Ch. | scroll | 26 | 135 |
| Ch.80.IV. | IOL Tib J 452 | 1. Suvarṇaprabhāsottamakīrtivarṇana; 2. Anuśayapraṇidhāna?; 3. Amitābhaguṇayogāsana (LVP 452) | Tib. | concertina | 9 | 30.5 |
| Ch.80.IV.b | IOL Tib J 463 | 1. Aṣṭadarśananāmamahāyānasūtra; 2. Aparimitāyurnāmamahāyānasūtra (LVP 463) | Tib. | concertina | 9 | 28.5 |
| Ch.80.IV.c | IOL Tib J 441 | Āryasarvabuddhaaṅgavatīnāmadhāraṇī | Tib. | pothi | 6.2 | 21 |
| Ch.80.IV.d | IOL Tib J 337 | 1. ĀryaCandanapratimāṅga; 2. Ākāśagarbhahṛdaya; 3. ĀryaAmṛtacakravidhi; 4. BhagavadAvalokiteśvarahṛdaya 5. Namaskāras (LVP 337) | Tib. | concertina | 4.7 | 17.8 |
| Ch.80.IV.f | IOL Tib J 460 | 1. Aṣṭadarśananāmasūtra; 2. ĀryaKanakavatīnāmadhāraṇī (LVP 460) | Tib. | concertina | 8.5 | 25.5 |
| Ch.80.IV.g | IOL Tib. 741.2 | Divination text (LVP 741) Saddharmapuṇḍarı̄kasūtra (Enoki C26) | Tib. | scroll | 26 | 135 |
| Ch.80.IV.h | IOL Tib J 748 | Prognostications for different years | Tib. | scroll | 26 | 96 |
| Ch.80.IV.4 | Or.8210/S.2554 | Commentary on a work of the Dharmalaksana School (Giles 5846) | Ch. | scroll | 29 | 1706.8 |
| Ch.80.IV.5 | Or.8210/S.5009 | Prajñāpāramitāsūtra (Giles 1549) | Ch. | scroll | 28.1 | 45.72 |
| Ch.80.IV.7 | Or.8210/S.5044 | Mahāprajñāpāramitāsūtra (Giles 25) | Ch. | scroll | 777.24 | |
| Ch.80.IV.12 | Or.8210/S.5013 | Saddharmapuṇḍarı̄kasūtra (Giles 2312) | Ch. | scroll | 25.7 | 840 |
| Ch.80.IV.13 | Or.8210/S.4640 | List of Buddhist works (Giles 7903) | Ch. | scroll | 30.7 | 29.4 |
We can see that the manuscripts in the above two bundles (B&C) are certainly not grouped together according to any Buddhist cataloguing system. Based on this initial study, there seem to be three possible conclusions: 1) Stein’s numbering is not consistent; 2) the contents of the bundles are not as originally placed in the cave; or 3) not all the bundles comprised contents selected according to a catalogue or other system. It is impossible to be certain about 1) or 2) but, as discussed above, the evidence we have goes against either of these conclusions. We explore 3) further hereafter.
Rong suggested that many bundles comprised manuscripts discarded by other monasteries or collected elsewhere to fill in gaps in the Sanjie Monastery collection. He traces the activities of Daozhen (道真), a tenth-century monk who was known to have made acquisitions for the monastery. Or.8310/S.6225 (Ch.75.V.3), for example, is a list of such texts with a handwritten note on back from Daozhen.66 It is certainly feasible that this bundle and some of others comprised manuscripts collected by Daozhen. Another possibility is that manuscripts were placed randomly in bundles once considered to be ‘finished with’: a form of sacred waste. But a further hypothesis worth exploring is that some were deliberate collections, possibly for example, made by individuals, whether monks or lay believers, and that perhaps they were placed in the cave after the owner’s death.67 Or perhaps the owner had died many years previously and the manuscript bundle kept for safekeeping elsewhere, for example, in a monastery library, until transferred to the cave. It is, of course, almost impossible to test this hypothesis especially with the very sparse samples to date but there is one bundle that provides some tantalising supporting evidence, although far too tentative at present to draw any conclusions.
The printed almanac, Or.8210/P.6, can be linked to a local astronomer and almanac writer, Zhai Fengda (c. 883–966). Several of his manuscript calendars are found in the Library Cave (Teiser 1994). The almanac survived as two incomplete parts which had been joined together on the back with paper fragments. As Giles discusses (1939: 1034), these contained text, including the end of a contract with the names of the landlord and two witnesses. Two other parts both mention a man named Zhai. One has a line in black ink reading ‘…written by Zhai, Group Head, Erudite/Doctor in the Prefectural School, and Vice-President of the Censorate, on the 26th day of the fourth month.’ The last character has been corrected in red and there is a shorter line following, also in red which reads: ‘For the attention of bhadanta Qu out of gratitude for his fatherly care.’ The second text consists of four lines reading: ‘Presented by city governor Zhai on the journey east… ‘. Zhai is almost certainly Zhai Fengda and it is a reasonable hypothesis that he acquired this almanac and reused paper with rough copies of notes he had made as backing to extend its life. The almanac probably dates to 877 and was not made locally. Zhai would certainly have been interested in it professionally for its content but also, possibly, by the fact it was printed: none of Zhai’s own calendars are printed. The Stein id4 on the almanac is Ch.91.I.3. The only other item with this bundle number currently recorded on the IDP database is IOL San 1446 (Ch.91.I.2). This is a sheet of paper containing stamps of Buddhist figures and Sanskrit dhāraṇī on one side. Stamps are, of course, also a form of printing.
One of Zhai’s manuscript calendars, probably for the year 956, was presented to the then-ruler of Dunhuang, Cao Yuanzhong (r.944–74) (Tesier 1994: 120–1).68 By this time printed texts were being produced at Dunhuang under Cao’s patronage. Several copies of a Buddhist prayer sheet survive, showing a Buddhist figure in the top half and text in the lower and commissioned by Cao on the occasion of the Ghost Festival in 947.69 The woodblock carver is named as Lei Yanmei.70 Two years later Cao commissioned Lei to produce a printed Diamond Sutra ‘to promote high moral standards’.71 Interestingly, this is a codex — more like a modern bound book — rather than scroll format. This might show Cao’s interest in changing technology: the codex form was being developed at this time. But it is also relatively small format, 14 × 10 cm, requiring a smaller woodblock and smaller sheet of paper than for printing a scroll (where each panel of paper was c.46 × 31–32 cm).72 The paper is quite coarse, almost certainly of local production, and the printing is nowhere near the quality of the Diamond Sutra of 868 (The British Library, Or.8210/P.6). Fine quality paper, which we know was often obtained from China and not necessarily locally produced, might have been in short supply or perhaps this production was deemed experimental and not worthy of using the finest paper. All this suggests that Dunhuang was starting to develop printing technology at this time but that it was still a long way behind Tang China: or indeed Korea and Japan where printing is seen from around 700.
As mentioned above, case 91 is problematic. If we look at Stein’s packing list for case 91, he notes that it contains ‘MSS, from Ch., 1 large Tib. bundle; 2 Chin. bundles; Tib. pothi.’ (MSS Stein 37/120) The almanac was presumably from one of the Chinese bundles. There is one other Chinese manuscript recorded with the id4, Ch.91.II., but these are the only 3 manuscripts so far recorded with this id. (Although, as mentioned above, some might have the id. Ch.9.) We will have to wait to see if more manuscripts/printed documents emerge. Until we have more data, the hypothesis that this bundle contained items associated with Zhai Fengda, if not in his possession, remains extremely tentative.
Conclusion
Based on an extensive examination of the documentation relating to the Library Cave, we show that Stein’s documentation is helpful in reconstructing the contents of some of the bundles as they were found in the cave in 1907 when Stein arrived. We further argue that it is probable that many of these regular bundles contained their original contents, although in some cases incomplete as some items might have been previously removed by Wang to send to local officials. From examination of a small — but random — sample, and building on Stein’s original description of four types of bundles, we suggest an at least one additional type, B5 below:
B1: Miscellaneous bundles of painted and textual material, often in languages other than Literary Sinitic and Tibetan, and some of it from sites other than the Library Cave.
B2: Miscellaneous bundles of paintings, textiles, banner staves etc, which comprised temple sacred ‘waste’ (eg. Ch.xxxv. and xxxvii.).
B3: Miscellaneous bundles of torn ends of sutra rolls; wooden rollers, silk tapes; cloth wrappers which comprised library ‘waste’.
B4: Regular bundles, probably mainly in hemp covers some of which were inscribed, containing Buddhist sutras, some catalogued according to the Kaiyuanlu.
B5: Groups of largely textual material in different categories, possibly including items collected by Daozhen—and maybe others—for the monastic holdings, individual collections or items brought together for other, non-random reasons, yet to be proposed.
The next step is to complete the recording of the Stein site numbers from the manuscripts and make these fully accessible with the other data held on the IDP database. It should then be possible to reconstruct more of the bundles and to test further these hypotheses and/or propose new ones. Stein was alert to this possibility, the reason he tried to document this material as comprehensively as possible in the very limited time available. Over a century later, and we are still to make full use of this important legacy to understand more about Wang Yuanlu’s exceptional discovery and to move beyond studying the manuscripts and their texts in isolation and place them in their Library context.
Notes
[1] Other Europeans had visited before 1900, including St George (1851–1931) and Teresa (1839–1928) Littledale in 1895; a Hungarian team, under Count Bela Széchenyi (1837–1918), who produced line drawings and written descriptions of the cave site after their visit in 1879 (see Russell-Smith, 2000); and, a few months later, Nikolai Mikhailovich Przhevalsky (1839–88) who made drawings of the site (1883). But these visits were all before the Library Cave was discovered. Lajos de Lóczy (1849–1920), a member of the Hungarian expedition, described the site to Stein in 1902, one reason for Stein’s visit. Charles-Eudes Bonin (1865–1929) visited in 1900 before the discovery of the Library Cave. He took rubbings of some of the Chinese inscriptions in the caves (Bonin (1901) and Doumy’s paper here). See also Enoki (1980).
Stein first went to the Buddhist site for a day on March 16, 1907, and looked at the corridor of Cave 16, but the door to cave 17 was ‘locked with a rough wooden door.’ (The Bodleian Library, MSS Stein 1862–1943: 294/267 (hereafter MSS Stein); Stein 1912, p.165, n.19). He did, however, see a Chinese Buddhist manuscript from the cave which was held in an adjacent modern Buddhist shrine (Stein 1921, 802).
[2] He also notes that this accords with Stein 1921: 802, 803–4.
[5] Correspondence between Stein and Chavannes concerning this dating to July and August 1917 is in MSS Stein 70/18–21 and 59–64. See Annick, forthcoming, for further discussion. A copy of Jiang’s original transcription is in MSS Stein 283/79. I have not found the rubbing: it was possibly sent to Chavannes in Paris for translation.
[6] He was allowed in later after Wang had removed most of the scrolls and he took measurements then for his plan (1921: Plan 4).
[7] This is the platform built for the statue of Hong Bian (洪辯 d.862): the cave was originally excavated as his memorial chapel, as shown by the inscription on the slab (Ma 1978). The statue is believed to have been removed many centuries before when the cave was originally filled with manuscripts. Again, there is no conclusive evidence for this. Of course, it is possible that this was the original position of the miscellaneous bundles and not one chosen by Wang.
[8] It is, of course, also possible that the cave was in fact sealed later than generally assumed and that this material was added on the top over the following centuries although, as Stein noted, the style of the painting over the hidden door challenges this.
[9] Also see Rong 1999–2000: 270, who suggested Cave 464 as the source of some of the material and his comments on 272 about other material.
[10] From Laurence Binyon’s letter to Stein, dated 7/3/1913. See also Pelliot 1908: 529 and Stein 1921: 828–9,
[11] For a clear example, see the painting now in the British Museum, 1919,0101,0.140, Ch.lii.001 (https://www.britishmuseum.org/collection/object/A_1919-0101-0-140). The discussion on this is summarized in Whitfield and Sims-Williams, 2004: cat 243, with further references. There is the further issue of forgeries. Forgeries had been produced before the cave’s discovery, notably by Islam Akhun and others from Khotan who sold their creations to the British and Russian consuls in Kashgar. See Sims-Williams and Waugh 2010 for further discussion. For Dunhuang forgeries see the papers in Whitfield 2002, especially Cohen, cited above, who potentially identifies one of the pre-Stein dispersal paintings as a forgery. However, since material acquired by Stein in 1907 is well-provenanced, they are not relevant to the argument presented here.
[12] Although this, of course, assumes that the ‘confusion prevailing’ was not as they were originally placed. But he thought not as he notes that: ‘Several small bundles of “waste stuff” examined yesterday & today were fully sewn up. This suggests that they were deposited at leisure, not at time of commotion & danger.’ (MSS Stein 204/319). Also 1912: 188.
[13] ‘The bundles were almost all sewn up tightly in coarse covers of linen.’ Stein elsewhere described them as canvas. Given linen is not found at Dunhuang and hemp is, we assume these were hemp. There are several hemp sutra wrappers from the Stein collection in the British Library, eg. Or.8210/S.11463, some undoubtedly used to wrap the ‘regular bundles’, but others of which might have come from bundles which Stein described as ‘waste’. See discussion below and thanks to Mélodie Doumy for locating and discussion on these.
[14] https://www.britishmuseum.org/collection/object/A_MAS-858 and https://www.britishmuseum.org/collection/object/A_MAS-859. ‘MAS’, Stein’s initials, is the prefix given by the British Museum to artefacts from Stein’s second expedition. Paintings and some textiles have the prefix 1907 to indicate the year of their registration into the British Museum collection, while Chinese manuscripts, which were part of a different department, have the prefix Or.8210 – ‘Or.’ indicating the ‘Oriental collections’ and 8210 the serial number assigned to Stein’s 2nd expedition. Stein material was also sent to India.
[15] If we assume 15 scrolls per bundle, even if he gave away most of the contents of each bundle he examined, then he still would have opened 20–50 bundles.
[16] A Stein himself notes, it was possible to examine and remove material from the wrappers with open ends leaving the remainder of the bundle intact. Stein found much of his non-Chinese material by doing this.
[17] The Tibetan and Chinese language material was probably almost equal in the space it took in the cave.
[18] There has been confusion over this narrative in some previously published accounts which is a reason to present a detailed chronology here along with the documentary support.
[19] The discrepancies are minor and most convincingly explained by the rushed nature of Stein’s activities over that period, rather than any attempt to mislead. The main narrative source is from MSS Stein 204–5, which are typed transcripts of his field diaries (the original diary for his 1907 time at Dunhuang is MSS Stein 199), but also MSS Stein 141–2, which are his pocket notebooks from February to October 190 and various letters to friends.
[20] Although on the top, they still required effort to retrieve as Stein notes: [Wang] ‘grew visibly tired climbing over manuscript heaps and dragging out heavy bundles’ (1912: 179).
[21] Stein describes Jiang’s camp as ‘at the very feet of a colossal seated Buddha reaching through three storeys, and with his innate sense of neatness promptly turned it into quite a cosy den with his camp rugs.’ (1912: 164–5).
[22] British Museum, MAS.0.1129 (https://www.britishmuseum.org/collection/object/A_MAS-0-1129). Given its fragmentary state, this might have well been one of the pieces which Stein mentions as having been placed on the floor.
[23] By ‘the house’ he means a hall in one of the two unoccupied houses next to the caves, which is ‘still possessed of a door and trellised windows’ (Stein 1912: 164).
[24] In his diary he says ’13 bundles of Chin. rolls (mostly intact).’ MSS Stein 204/339. In his 1912 and 1921 accounts he says ’20 bundles’ (193/825).
[25] He gives these figures in his official report on his second expedition submitted to the government of India on 13 November 1908. He asked for the section on Dunhuang to be kept confidential in case he had further opportunity to acquire more material from Dunhuang (Wang 2004: CE32/23/16/2]. Stein’s diaries note 12 cases of manuscripts and 4 of paintings. (MSS Stein 204/344–45). His later published works give slightly different figures: eg in Desert Cathay (1912 II:193) he writes: ‘He agreed to let me have fifty well-preserved bundles of Chinese text rolls and five of Tibetan ones, besides all my previous selections from the miscellaneous bundles.’ […] we succeeded in making him stretch a point further, and allow me to add some twenty more bundles of manuscripts to my previous selections,’ ie 75 bundles plus previous selections. But he did not always have all his original documentation to hand so these differences in figures are not surprising.
[26] AS to Percy Allen (hereafter PA), 14 October 1907. Also letter to Fred Andrews (hereafter FA), 230 bundles 1 October 1907 (MS Stein 37/104). And in his report: ‘acquisition of 220 more bundles of manuscripts just received at An-shi. October 1. The number of texts is raised thereby to close on 4.000’. (Wang 2004: CE32/23/16/2.) His diary for 5 Oct.: ‘’Last night [ie Oct 4] Ibrahim Beg arrived at 10 PM with 4 camels bringing purchases from Tao-shi, 230 bundles (among them 20 Tibetan). Convoy had marched all night to avoid attention. 7 big Taghars [bags] full of MSS. were unloaded behind temple & warily brought in by Tila & Ibraham between 10–11 PM.’ (MSS. Stein 205/94–5). Stein (1921) also gives 230 (825).
[27] ‘Settled business at last by adding the extra weight in bullion. 600 odd rolls were to be delivered, and no payment asked for “Pusas”.’
[28] ‘At Tun-huang it was not long before some unknown Chinese well-wisher presented himself with a fairly large packet of manuscript rolls from the same source which he was anxious to dispose of. Judging from the very modest compensation which induced him to return later with more, I was able to conclude that the article was not altogether a rarity in the local market.’ (1928: 355). Also see MSS Stein 216/21 and 22, dated 1/4/1914 and 2/4/1914 and, confirming the purchase, Stein’s account books for these dates are at MSS Stein 152/24.
[29] ‘at Kashgar and Khotan, through scattered rolls of Buddhist chings, manifestly derived from the Ch’ien-fo-t’ung hoard, which had found their way in the hands of Chinese officials and in a few instances had been presented to George Macartney and others.’ (Stein 1928: 355); and again ‘numerous specimens of such rolls … which were subsequently shown to me at different Ya-mens along the route.’ (356).
[30] ‘The temples have remained places of worship for the pious of Tunhuang down to the present day, and the removal from them of any fresco paintings of sculptures could not have been thought of even if technically practicable.’ Letter from Stein to Secretary to the Government of India, Home Dept. 26/9/1907, Wang 2004: CE32/23/10.
[31] It is regrettable that the site id, when given, was not always recorded in catalogues and databases of the material. A notable omission is in Giles’s catalogue of the Chinese manuscripts (Giles 1957). IDP started recording this information on its database at the British Library, but the work has not yet been completed.
[32] ‘I may note here that when the marking with serial numbers was made at the British Museum …’ (1921: 814 n.2).
[34] 1921: Appendix F gives a listing of the manuscripts.
[35] ‘Only in a few cases, e.g. Ch.0079. a, b; 00275, have Brahmi texts been described without an indication of the “mixed” bundles in which they had originally come to light.’ (1921: 814 n.2).
[36] The photograph usually reproduced (as in Stein, 1912: fig 188 and Rong, 2013), is a forgery, although with no intent to deceive. Stein took an image but it was double exposed, as shown here. He therefore recreated it later by drawing the manuscripts on an image of the empty Cave 16. The original double-exposed print and the forgery are in the British Library collections (Photo 392/59(1 & 2)).
[37] The photograph shown here has hundreds of bundles but not over 1000; also no visible pothi, so there must have been more material piled in the corridor outside the shot.
[38] ‘The number of individual rolls in the 270 of regular bundles which I secured in the end in addition to previous “selections” proved so great that, when at last in July, 1908, there was leisure to set Chiang Ssu-yeh to work on them, the weeks available before my final departure from Khotan sufficed only for a first rapid listing of less than a third of them.’ (1921: 916). The mention of 270 bundles presumably refers to those acquired in Ac1d and Ac2. Given that the number of bundles in Ac2 is variously given as 220 or 230 bundles, then we can assume that Ac1d contained 40–50 bundles.
Also, Stein to PA, 23/6/1908: Jiang ‘half smothered under piles of text rolls and bundles of “Misc. MSS.”…’ (MSS Stein 5/122–3). ‘catalogued about a third, to over no. 1100.’ (MSS Stein 5/128, 6/7/1908).
[39] ‘Pelliot at work on Chinese MSS. agreeing to take up inventory.’ AS to PA, 1 July 1910. MSS Stein 7/72–3.
[40] The mazi numbers currently recorded include 1301 (Or.8210/S.554) which would accord with Stein’s note in his letter cited above. A few later mazi numbers are recorded but, from checking a few, these are probably errors in reading the mazi and need to be corrected. Many of the mazi numbers are transcribed in blue pencil in Arabic numbers. There are also Chinese manuscripts without mazi but with Arabic numbers in blue pencil in the 1500s and 1600s. A reasonable hypothesis would be that these are a continuation of the mazi sequence made in London, but then we would hope to find them in 1300s and 1400s. Or they could be misreadings of the original mazi. A further complicating factor is that blue pencil Arabic numbers are also found on manuscripts without mazi, but this could be that the mazi were written on fragments on the manuscripts which have since been removed. Until more data is available, it is not possible to test these hypotheses.
[41] Giles helpfully published a series of articles of the dated manuscripts in the Stein collection (Giles 1935a, 1935b, 1937, 1940, 1943). See http://idp.bl.uk/database/oo_loader.a4d?pm=Or.8210/S.316 for the image.
[42] eg Or.8210/S.80, noted as Ch.480 in Stein 1921: 672. Importantly, of the checked manuscripts, to date no other has contained this mazi number.
[43] This copy contains annotations by Andrews from 1910. There is an earlier copy at 155–8 without these.
[44] Or.8210/S.88 has mazi 6 in red ink. According to the packing list, this manuscript would have been in case 67. The Id5 is unclear and not in usual format, as there is no Ch. but the number 74 (or possibly 76) with a line below and, below that CLXVII OR ChXVIII.
[45] Report of Laurence Binyon to BM Trustees 5 August 1914 in Archives of the Oriental Dept ORIS 1913–16, 44.
[46] ‘Thomas called last week and it was arranged between him and Barnett that Thomas should examine the Tibetan manuscripts and “unknown’ scripts. Accordingly, yesterday I made up a small packet of Tibetan mss. for the IO [India Office] messenger.’ MSS Stein 37/219, letter from FA to Stein, dated 14/11/1909.
[47] Caroline Ridding’s contribution is mentioned in the preface. See Huett 2012 for her work on manuscripts from the Younghusband expedition. She is acknowledged as joint author on the University Library of Cambridge catalogue of Sanskrit manuscripts: they completed this together in 1916.
[48] It is implied in the introduction to the catalogue: thanks to Sam van Schaik for this observation.
[49] This separation of the manuscripts between the BM and the IOL was discussed and, in October 1914, Barnett proposed that the British Museum retain most of the Chinese manuscripts while all the Tibetan manuscripts were ‘assigned to the share of the Secretary of State for India.’ Wang 2004: CE32/23/57, letter from Thomas to Kenyon, dated 21 October, 1914. Thanks to Sam van Schaik for information on the index cards.
[50] As Sam van Schaik points out, the ids in the catalogue were the result of further work by Assistant Keeper Miss A. F. Thompson who compiled the index. van Schaik suggests that she might have been responsible for adding and editing the ids before publication in 1960. Also see van Schaik 2007.
[51] ‘Only in a few cases, e.g. Ch.0079. a, b; 00275, have Brahmi texts been described without an indication of the “mixed” bundles in which they had originally come to light.’
[52] For examples, see British Museum, 1928,1022.24 (Ch.05) and The National Museum, New Delhi, 2003/1875 (Ch.024).
[54] As argued above, the issue of possible forgeries from Stein’s third expedition is not relevant to the discussion but, for Fujieda’s argument (2002) and other papers in Whitfield 2002.
[55] [Wang] now produced two big boxes crammed with well-preserved manuscript rolls. By the careful appearance of their writing and the superior quality of the paper it was easy to recognize that the specimens I was able rapidly to examine belonged to that great stock of canonical texts, mostly Buddhist…’ (1928: 358).
[56] Stein normally distinguished purchased–and thus less well provenanced—material from that which he excavated, but since, in this case, acquisitions from both Wang and others were, in effect, not directly seen to be from the library cave, he might well have used a single system.
[57] Stein notes that the bundles containing the 570 scrolls showed evidence of being opened by Pelliot which would accord with Pelliot’s assertion that he examined everything in the cave [MSS Stein 216/28].
[58] with letter from Binyon dated 3 February 1915 and reply from Ross (War Office) on 26 Feb. 1915 agreeing to modifications. Binyon points out some items missing from Petrucci’s report, but these do not include Ch.i.001.
[59] It might not be the only one. Binyon’s correspondence with Stein from October 1927 notes: ‘And in fact on each occasion when the collection has been handed over some items have been reported missing thought they usually reappear.’ Whitfield will write further on this in her blog post: SilkRoaddigressions.com.
[60] T12, no. 360, also known in English as the Larger Sutra on Amitāyus, the title used by Giles.It was translated into Chinese in the middle of the 3rd century by Saṃghavarman / Kang Sengkai 康僧鎧.
[61] Fujieda Akira (2002) and J.P. Drege (2002) were among the first to note the importance of the paper, type and size, in grouping and dating manuscripts. As Fujieda notes, 31–32 cm is the height of paper used for official documents prior to the Tibetan period in Dunhuang after which it starts to be used for Buddhist manuscripts as well: Buddhist texts are previously mainly on paper which is 26–28 cm high (2002: 104).
[62] Wang 2004: CE32/23/57 Letter from Barnett to Kenyon dated 21 October 1914.
[63] Also see Rong (1999–2000: 253). The sutra is the forty fascicle version of the Mahāprajñāpāramitāsūtra. The Kaiyuanlu catalogue used characters from the Chinese text, Qianzijing (The Thousand Character Classic) as a short identification of the catalogue sections.
[64] The existence of apocryphal sutra in the bundle is not evidence, however, that this did not belong to a Buddhist monastery, as Rong discusses. He notes that ‘in the tenth century, many apocryphal sutras circulated at Dunhuang.’ (1999–2000: 262).
[65] The Buddhist community might also have reused lay documents for copying practice, but the dates of the club circular suggest the sutra was written first.
[66] We have looked at other items in the Ch.75.V. bundle, but only a few are recorded and they are at the end of the Or.8210/S. sequence which comprises fragments of manuscripts only catalogued from the 1980s (Rong 1998). Many of these ‘fragments’ were in fact pieces which had originally been pasted onto other manuscripts and removed by conservators.
[69] The British Library, Or.8210/P.8, Or.8210/P.9 (Lei Yanmei mentioned in the latter); Le Bibliothèque nationale de France, Pelliot chinois 4514 (Rong, 2004: 62). As Giles points out, the Chinese year name used in this text ‘4th year of Kaiyuan of the Great Jin dynasty’ was incorrect, the dynasty having changed in the 2nd month, but the news had not reached Dunhuang (Giles, 1943: 149). It is worth noting that Or.8210/P.9, like P.6, is pasted onto other paper, possibly to extend its life. Some other Dunhuang manuscripts present the same kind of discrepancies, but others from the same years are correct (Arrault and Martzloff 2003: 93).
[70] As Giles points out, he is called the ‘artificer’ in this first text but, two years later, has the official title ‘ya-ya’ (1943: 152).
[71] The British Library, Or.8210/P.11; Le Bibliothèque nationale de France, Pelliot chinois 4515 and 4516 (Rong 2004: 62).
[72] Fujieda 2002: 104.
Funding Information
Paschalia Terzi carried out this work by getting an Erasmus training placement grant.
Competing Interests
The authors have no competing interests to declare.