In the 1770s the Qianlong emperor set out to do something no Qing ruler had attempted: to take possession of the written record of his own empire, down to the last commentary on the last classic, and to keep it inside the palace walls. The instrument was a library, the Siku Quanshu, the Complete Library of the Four Treasuries, assembled by a state apparatus of scholars, scribes and officials over more than a decade and copied out seven times. It was also, in the same breath, a censorship machine, because an empire that claims to hold everything must also decide what may not be held.
What follows is the story of that double enterprise as an archival one. A library had to be gathered from private owners who were told they could have their books back, copied by hand onto folded paper in a rigid format, bound in coloured silk, described in a catalogue, and distributed among seven specially built halls so that it would outlive the man who ordered it. Running alongside it was a second apparatus of inspection and destruction that kept its own register of condemned books. The registers have survived better than the objects, which is why we can still say what was saved with more confidence than we can say what was burned.
The Emperor Who Wanted the Written Record
The Qing court had no reason to doubt that it would outlast its enemies. What it could not assume was that the texts which defined its own legitimacy would remain legible, complete and uncontested for as long as the dynasty lasted. The Manchu rulers had taken a Chinese throne, and a throne of that kind had to be argued for in writing, using texts that had been in Chinese hands for a thousand years. Copying those texts into an imperial collection was therefore not housekeeping. It was an act of sovereignty, and the catalogue that described them was the argument itself: this is what the empire contains, arranged in an order the emperor approves.
The collection began with a sweep rather than a library. In 1772 the Qianlong emperor ordered local and provincial officers across the empire to search private and official holdings for books worth copying, and to send them to the capital. The message offered to the owners was reassuring in the way such messages usually are. Books submitted would be returned after copying, and no punishment would follow from surrendering them. Thousands of volumes duly arrived. Accounts differ on the exact terms, but the scale of the response is not in doubt: provincial officials shipped crates of books towards Beijing, and the Qing state was operating a compulsory survey of the national literature in everything but name.
A formal decree followed, dated to 1773 in the standard reference accounts, establishing a dedicated bureau, the Siku quanshu guan, whose only task was the compilation. The bureau had a hierarchy of chief ministers who supervised it, chief compilers who directed the editorial work, a division for collation, a division for the copying itself, and a division for the manufacture of cases and furniture. The subjects were divided into four divisions, the four treasuries of the title: the Classics, the Histories, the Masters or philosophical writings, and the Collections, which held literature, poetry and the arts of the brush.
How large the result was depends on which set of figures you consult, and the honest answer is that the traditional numbers are traditional rather than checked. The most commonly repeated account gives 3,461 titles in 793 chapters and 35,940 scrolls, with a full catalogue running to more than ten thousand entries. The libraries that describe their own copies give slightly different figures, since the same project can be measured in titles, chapters, scrolls, bound volumes and characters. The HKBU Libraries description of its Yinyin Wenyuange edition speaks of about 3,500 individual titles in roughly 80,000 volumes and around 800 million characters; the standard summary of the collection counts close to a billion words and roughly 36,000 bound volumes for each set. These are not contradictions so much as the rounding of different measuring sticks applied to an enormous object.
Even so, the scale is the point. This was a library assembled by thousands of people over a decade and a half, copied seven times in full, and then described volume by volume in a catalogue that runs to hundreds of chapters. As R. Kent Guy argues in his standard monograph, the Siku Quanshu was less a library in the modern sense than a state operation for fixing the boundaries of the written past, with the catalogue as its product. What the emperor wanted was not a collection but a description: a map of the literature of the empire, drawn by the empire, in which every included work had been collated, approved and given a place.
That ambition carried a second edge, and the sources are frank about it. A project that inspects every library in the empire in order to copy what is good must also decide what is not good. The same officers who searched the shelves were expected to report on what they found there. The catalogue that recorded what the empire owned would therefore have a shadow catalogue recording what the empire rejected, and that shadow catalogue was not a marginal by-product. It was a working instrument, used to destroy books across the country.
Collecting: Requisition, Scholars and Scribes
The machinery of collection was a transaction in which the state held all the leverage. Owners were asked to hand over books and told they would get them back; in practice some did, many did not, and a substantial number were never returned. Once a book entered the bureau it was exposed to the full editorial process, and it might be corrected, abridged, described without being copied, or refused outright. The scholarly history of the collection is in large part a history of owners trying to recover property deposited as a favour and treated as a contribution.
The editorial team assembled to do the work was among the most formidable concentrations of Chinese scholarship in the eighteenth century. The chief compilers were Ji Yun and Lu Xixiong, both of whom had already made their names in textual criticism and would make them again as editors and commentators. Around them worked a roster that includes Dai Zhen, the philosopher and philologist whose evidential method demanded exactly the close attention to wording that collation required; Shao Jinhan, whose editions of ancient texts became standard; and Zhou Yongnian, Weng Fanggang and Yao Nai of the northern school of learning, which treated the received texts as objects of study rather than of reverence. Supervision belonged to senior ministers including Yu Minzhong, Peng Yuanrui and Zhu Gui, which is a reminder that the project was judged at the highest level of the court.
The technical work was what made the collection authoritative. For each candidate work, the editors assembled multiple copies, often from different provincial libraries, and compared them character by character to establish a text. Where copies disagreed, the reading was supported by the older manuscript traditions, by parallel passages in other works and by the judgement of the scholar responsible, and the procedure was recorded in the catalogue and in the preface for each class. The result was not a reproduction of what libraries happened to hold in the 1770s but an authorised edition, a fixed text, with a documented argument for why it was the right one. Where a work could not be established beyond reasonable doubt, it was described but not copied, and the description alone became a record of what the book had contained.
The copying itself was an industrial operation carried out by hand. The number of scribes usually given is about 3,800, and another count puts the number of copyists at 3,826; the workforce of scholars and examiners behind them is given as more than 3,600, which probably counts the same men at different stages of the work. They were not, in the usual account, paid in cash. A copyist who finished a prescribed number of sections could expect an official post, so that the labour was rewarded with places in the bureaucracy and the work acquired the character of an examination in itself. Every volume was written in a clear regular hand, the kaishu script, on folded paper printed with a red frame, each page divided into two halves and each half into eight columns of twenty-one characters. The rules made the page a grid, and the grid made error control possible: a misplaced character in a column of regular squares is visible at a glance.
The system also kept a paper trail inside the book. Names of the copyist and of the checker were entered at the end of each volume, and during a later checking pass yellow slips were pasted inside the front cover, so that a fault found in a second collation could be run back to the hands and the volume that produced it. The physical form was equally prescribed. Each text was issued as a thread-bound booklet, the leaves folded and sewn at the fold in the Chinese manner, and the booklets were then enclosed in cases. The covers were sewn in coloured silk, and the colour told the reader where the book belonged: green for the Classics, red for the Histories, blue for the Masters, grey for the Collections. A shelf of the finished collection therefore looked like a code, with bands of green, red, blue and grey running across the cabinets, and a librarian could see at a glance whether a volume was in the right place.
Seven copies were planned, and the first was finished in 1782 and placed in the Wenyuan Pavilion in the Forbidden City. The other northern sets followed for the palaces at Chengde and Shenyang and for the Old Summer Palace, while three further sets were ordered for the south, at Zhenjiang, Yangzhou and Hangzhou, where local merchants paid to build and furnish the halls that would hold them. A reserve copy was also kept in the Hanlin Academy in Beijing. The first copy is generally given as complete in 1782, the last deliveries and the final revision of the catalogue running into the early 1790s, and the National Palace Museum in Taipei gives 1787 as the year the copying programme concluded. The last of the seven was already in its hall before the Qianlong emperor died in 1795.
The Catalogue That Decided What Would Live
The great catalogue was the Siku Quanshu zongmu tiyao, the classified summary of the complete library, two hundred chapters long, finished in 1781 and printed in 1793. It is the single most important document to come out of the project, more consequential than any of the sets themselves, because the sets are copies of books while the catalogue is an argument about which books matter. Ulrich Theobald’s reference account sets out its scope precisely: the summary describes 10,254 books in 172,860 scrolls, of which only 3,461 books in 79,309 scrolls were actually copied into the collection, while a further 6,793 books, running to 93,551 scrolls, are recorded without being copied. Other figures circulate, such as a catalogue of more than 10,680 titles, and the discrepancy is a matter of what counts as a title.
What the catalogue does with those books is the intellectual achievement of the enterprise. The four divisions are broken into forty-four sub-categories, and those in turn into sixty-six finer classes, a taxonomy that was new and deliberately exhaustive, extending to kinds of writing that had no place in an earlier canon, among them the collected works of painters, calligraphers and local worthies. Within each category the books are arranged in chronological order, so that the collection reads as a history of a subject rather than an alphabetical convenience. Each class opens with a preface discussing what the class contains and how the editors judged it, each division opens with a general preface, and the whole is fronted by the imperial edict, the list of compilers, the formal dedication and the rules of compilation, the fanli, which spell out what was to be collected, what excluded and how the work was to be judged.
The most revealing part of the catalogue is the category of books that were examined and not copied. The compilers called them cunmu, books available and checked, and for each they supplied a bibliographic description, often with an evaluation. In effect the catalogue preserves an abstract of thousands of works that the empire deliberately declined to reproduce, including works that were later suppressed and destroyed. For the modern scholar this is a gift and a puzzle. The library holds the texts that passed; the catalogue holds descriptions of the texts that did not, at a level of detail sufficient to show what the collection would have looked like had the exclusions gone the other way. Any history of Chinese bibliography written from the Siku Quanshu therefore rests on both halves at once.
Because the full summary is unwieldy, the compilers also produced an abbreviated finding aid, the Siku Quanshu jianming mulu, twenty chapters long, printed in Hangzhou in 1784 by Zhao Huaiyu. Lee Sang Yong’s bibliographical study of this short catalogue shows how closely its format was examined and corrected against the longer work, and how the two are meant to be read together: the short list tells a reader where a book is, the long summary says what the book is and why it was accepted. The short catalogue became the practical instrument for later generations, and the preface William Hung wrote in 1939 for his index to the summary and to the companion catalogue of books not received shows how completely the compilers’ ordering became the reference system of the field.
Not everything was meant to fill a shelf. A selected body of 473 works was extracted into a portable summary, the Siku quanshu huiyao, compiled by Cinggui and later printed as part of the Songlin collectanea, giving the emperor a condensed library that could travel with him and be consulted without handling thirty-six thousand volumes. The imperial catalogue, the huiyao, and the classified summary served different needs: one for the court’s daily use, one for the specialist, one for the repository. That system was in place before most of the books had been copied.
The catalogue has been corrected, supplemented and reprinted for two centuries, and the corrections are themselves a measure of the original achievement. The 1965 Zhonghua Shuju edition, based on the Hangzhou printing, added two documents that the compilers had never published: descriptions of the books eliminated or annihilated during the work, the Siku chehui shu tiyao, and Ruan Yuan’s later account of works that the compilers never checked, the Siku weishou shu tiyao. Later scholars added their own corrections, among them Yu Jiaxi and, after him, Hu Yujin and Wang Xinfu. The result is an apparatus of considerable scale resting on the original, and the summary remains the standard starting point for the study of almost any text written in China before the nineteenth century.
The Other Catalogue: The Books Marked for Suppression
Every book that failed the inspection had a fate, and the state kept the list. The register of condemned books is generally known as the Siku quanshu Kanding, the catalogue of forbidden books of the complete library, and the standard account of the project credits it with more than 2,855 titles rejected and banned during the compilation, plus several hundred more that were edited into acceptability rather than refused outright. The mechanism was simple and effective. A book that was not admitted to the collection and was not returned to its owner was liable to be entered on the list, and a book on the list was a book to be destroyed. The promise of return that had drawn so much material into the capital proved to be the weakest part of the bargain precisely where it mattered most.
Alongside the requisition ran a campaign of inspection and destruction that extended well beyond the books actually copied. The standard survey of the literary inquisition under the Qianlong emperor records fifty-three cases of literary persecution during his reign and describes a purge effort running from 1772 to 1793. It puts the number of works lost at as many as 3,000, and the number of volumes destroyed at an estimated 151,723. These are estimates and should be treated as such, since the surviving paperwork is partial and local officials had every reason to report large numbers of volumes destroyed and few titles examined. What can be said with confidence is that the scale of destruction was measured in tens of thousands of volumes, that it was carried out by local authorities under central direction, and that it continued for more than two decades after the collection itself was finished.
What was targeted has to be stated carefully, because the categories overlap. The obvious target was writing that could be read as disloyal: histories of the Ming dynasty, private histories written outside official auspices, works praising the dynasty that had just fallen, texts with anti-Manchu content, poetry that could be construed as subversive. Alongside these sat the apparatus of heterodox learning that the evidential scholars had spent a century displacing, and the unofficial literature of the late Ming, including novels and plays whose circulation the court distrusted for reasons that had little to do with politics. Local genealogies and gazetteers were vulnerable too, because they recorded names and histories that could be read in ways the compilers did not like. The system also had a self-interested local dimension, since provincial officials and private enemies could exploit the machinery of accusation for their own purposes.
The result is a paradox that anyone who has worked with the sources finds difficult to square with the image of a systematic purge. The compilers’ own later assessment, as it has been summarised, is that most of the works that were explicitly banned have in fact survived, while most of the works that were actually lost were never on the prohibited list at all. If that is right, the burning did its damage less through the official lists than through the ordinary chaos of the process: local rivalries, informants, the collapse of private libraries under pressure, the failure to return submitted books, and the simple fact that a book handed to the state in the 1770s was a book that often never went home. The list was the mechanism of record; the real losses were dispersed and untidy.
It is worth pausing on the fact that the list survived. The register of forbidden books, along with the descriptions of eliminated works that the compilers collected for the catalogue, is one of the most valuable documents in the history of Chinese censorship, not because it indicts the censorship but because it makes it legible. It tells a modern reader which books the court wished to see destroyed, which is a map of what the court feared. It also makes recovery possible, because a book known to have been burned can be traced in other libraries, and in a number of cases destroyed or damaged texts were found again in private hands and reprinted. The 1965 edition that added the chehui tiyao to the catalogue was in part an attempt to reassemble, in description, a collection that had been deliberately broken up.
The repression of books was also a legal matter, and it ended as one. In 1799, the Jiaqing emperor, the son and successor of Qianlong, declared that treating literary cases with the severity reserved for treason was unjust and inappropriate, and ordered earlier cases to be reviewed. The practice had run for close to a century and a half across the reigns of Kangxi, Yongzheng and Qianlong. The ending is a reminder of how contingent the whole apparatus was: it depended on the will of one man, and when he stopped applying it, the burning stopped. The archive of the Siku Quanshu therefore preserves two registers in the same imperial hand, one listing what the empire wished to own and one listing what it wished to destroy, and the second register is now as useful to historians as the first.
Seven Halls and a Painted Pavilion
The copies were not shelved in an abstract repository. Each was assigned to a building, and the buildings were chosen to say something. Four went to the north, to the emperor’s own residences: the Wenyuan Pavilion in the Forbidden City, built in 1775; the Wenyuan Pavilion at the Old Summer Palace, built in 1774; the Wensu Pavilion at the palace in Shenyang, built in 1782; and the Wenjin Pavilion in the mountain resort at Chengde, begun in 1773 and finished in the mid-1770s. Three went to the south, to Hangzhou, Zhenjiang and Yangzhou, in halls built with local merchant money and named by the emperor himself, the Wenlan, Wenzong and Wenhui pavilions. The northern four were the palace halls, where the emperor could consult the books himself; the southern three were placed where scholars and provincial candidates could reach them, which was a calculated act of generosity by a ruler whose name was to be carved on the volumes inside.
The architecture of these halls has been studied in detail, and it is not incidental. The obvious model was the Tianyi Ge, the old Ningbo library founded in the sixteenth century, and the seven halls were planned as imperial versions of a scholar’s library, with a two-storey central building, surrounding rooms and corridors, and painted decoration of the kind that marked a building in the imperial repertoire. A recent study by Yu, Liu and Yang treats the seven libraries as a single design problem, showing that the halls borrowed from the Tianyi Ge in plan, in their painted beams and in their furniture, and that the Wenjin Pavilion at Chengde is the most monumental of them, a two-storey structure built to look permanent because it was meant to outlast the dynasty.
Inside, the library was a piece of joinery and storage engineering. The Library of Congress record for the Wenjin edition describes the set as a manuscript of the Qianlong reign, comprising 36,304 volumes in 6,144 boxes, which is a useful reminder of what the shelves looked like: not open shelves of books but rows of cases, each case holding a fixed run of thread-bound booklets, each case labelled and ordered. The Qianlong emperor anticipated the long gap before the books would arrive by ordering imitations of the cases from an earlier imperial compilation to be placed in the empty cabinets, so that the new halls would appear to be full while the copyists were still at work. The effect was theatrical, and it was also a way of making the point that the missing books were not missing for ever.
Storage was a constant negotiation with fire, damp, insects and theft, and the cases, the silk bindings and the cabinets were all part of the answer. A thread-bound booklet in a fitted case could be lifted out, carried, inspected and put back, and a whole case could be moved in an emergency. The silk covers in four colours made it possible to see at a glance whether a volume had been returned to the right sequence. The court did not think of this in the language of conservation, but the collection was stored as discrete, movable, identifiable and replaceable units, which is precisely the logic that allowed parts of it to be rescued when the building around it was destroyed.
The southern halls were a different proposition altogether. The Wenlan Pavilion was built in Hangzhou in 1782 and finished the following year, beside a travel palace on the shore of the West Lake, and it survives as a two-storey library building set in a garden of water, rock and corridor. That setting was not decorative. A library placed in a garden, in the former capital of the Southern Song, was a claim about how the collection should be used, and the hall was built with private money and a promise of access for local scholars. The emperor’s words in the documents supporting the southern copies were widely understood to mean that men of letters in the region could consult the books on the spot. Whatever the practice turned out to be, the architecture made the promise visible, and it is the only one of the seven halls still standing in something close to its original form.
What Survived: Three Sets and a Half
The seven copies were distributed across a century of war and fire, and three sets were destroyed outright. The copy at the Old Summer Palace went with the palace itself when British and French troops burned the site in 1860. The Zhenjiang set was damaged during the war of 1842 and then destroyed by the Taiping forces in 1853; the Yangzhou set was destroyed the following year. The Hangzhou set was the worst hit of the survivors, with the standard account putting the losses in the Taiping occupation at between seventy and eighty per cent of the collection, though local bibliophiles began buying back scattered volumes as soon as they could. The four surviving sets all took damage during the war years of the late 1930s and early 1940s, when collections were moved inland and stored in caves. Separately, the destruction of books in Beijing during the siege of 1900, studied by Donald Davis and Cheng Huanwen, scattered the holdings of the Hanlin Academy, where a reserve set had been kept.
The first of the survivors is the Wenyuan set from the Forbidden City, now in the National Palace Museum in Taipei. It is the great manuscript of the collection, the model from which the other six were copied, and it travelled south with the palace administration after 1949. Its survival is the reason the collection is known at all in its original form: without it, the surviving sets would be copies of copies, and the physical evidence of how the eighteenth century bound and cased its books would be gone.
The second is the Wenjin set from Chengde, now in the National Library of China in Beijing. It was moved to the capital in 1914 and formally received by the new national library in 1915, and it is generally described as the only one of the seven that still stands in its original cases, on its original shelves, with its original arrangement. The Library of Congress catalogue describes the Wenjin edition as a manuscript of the Qianlong reign, 36,304 volumes in 6,144 boxes, and it is the set that has most often been treated as the reference copy for the physical condition of the whole project.
The third survivor is the Wensu set from Shenyang, and its history in the twentieth century is the strangest of the four. It was kept in the Liaoning Library in the early years of the People’s Republic, and in 1966 it was moved by train from Shenyang to Lanzhou for safety, as the Sino-Soviet split and the political climate made the north-east unsafe for anything so valuable and so ideologically marked. The volumes spent decades in Gansu, and they are now held by the Gansu Provincial Library in Lanzhou, where the anniversary of the library’s founding was marked in 2016 by an exhibition of original volumes and a commemorative volume on the collection. The irony is exact. The set that spent the Cold War in the desert north-west is the one that came through the twentieth century essentially intact, because it was moved out of the path of history rather than kept in it.
The fourth, at Hangzhou, is the small plain copy made for the Wenlan Pavilion, and it is the set with the most tangled rescue. Two Hangzhou bibliophiles, Ding Shen and Ding Bing, tracked down and bought back scattered volumes in the years after the Taiping occupation, and hand copies were then made of what could not be recovered, a work that continued into the Republican era and was completed by Qian Xun, founding director of the Zhejiang Library, and Zhang Zongxiang, the provincial education commissioner. The set was moved first to Guiyang and then to Chongqing during the Second Sino-Japanese War, and it now rests in the Zhejiang Library in Hangzhou.
The physical forms in which the collection survives are worth separating. There is the great manuscript of the Forbidden City, written out in full in the standard format. There is the Chengde set, variously described in the literature as a hand copy and as the printed set, with the Library of Congress treating it as a manuscript while other accounts stress the printed issue; the safest statement is that all seven sets were begun as hand copies in the same red-framed, eight-column format, and that any printed version of the Siku Quanshu belongs to the modern reproduction era rather than to the original seven. And there is the smaller plain copy from Hangzhou, made on reduced leaves, cheaper and more portable, the set that was scattered and then pieced back together by people who knew exactly what was missing. Completeness is a matter of degree, and the assessments differ: one scholarly account counts only three and a half surviving sets, while the Berkeley library that received the Wenlan reprint calls that set the most complete of the extant copies. Both judgements can be defended, which is itself a warning against confident statements about what survives.
The twentieth century then did what the Qianlong emperor could not, which was to make the collection portable. The Commercial Press began a facsimile series in 1935, reprinting selected works from the Wenyuan copy in thousands of volumes, and the series ran on for half a century. In 1986 the whole Wenyuan set was reproduced photographically in an edition of about 1,500 volumes, with the classified summary, an index and volumes of corrigenda, and digitised versions of that edition are now in wide use. The Hangzhou set followed in a reprint begun in 2004 and completed in 2015 in 1,559 volumes, of which three hundred sets were printed, one of them now in California. Electronic versions have followed, including the database published by Digital Heritage Publishing with the Chinese University of Hong Kong, which carries the complete contents of the Wenyuan edition in full text and image, over 800 million characters under almost two million piece titles, and which warns that the editors of the original were selective. Ronald Egan’s reflections on the uses of the electronic Siku Quanshu, published in 2001, caught the point early: a digital collection is not a photocopy of a library but a new instrument, and it can be interrogated in ways the shelves never allowed.
It is tempting to end with the loss, and the numbers make it easy: an estimated 151,723 volumes destroyed, a register of more than 2,855 condemned titles, three complete sets burned, a fourth largely scattered. But the more instructive figure is the one that survives the burning. Three thousand titles made the collection; some ten thousand more were examined and described; thousands more were destroyed without ever appearing on a list. The Siku Quanshu preserved a canon the Qing state could control because it had first destroyed what it could not, and yet the project also preserved, in its catalogue, an abstract of the excluded that no burning could remove. An empire that wanted to own its own literature ended by producing two archives: one of what it decided to keep, and one of what it decided to lose. The second archive, never meant to survive, turned out to be one of the most reliable witnesses we have to the shape of the literature the first could not admit.
References
- Siku quanshu (Complete books of the Four Storehouses) — Ulrich Theobald, ChinaKnowledge.de, 2010
- Siku Quanshu — Wikipedia
- The C. V. Starr East Asian Library Receives Monumental Gift — Institute of East Asian Studies, University of California, Berkeley, 2018
- Yingyin Wenyuange Siku Quanshu — HKBU Libraries, The University of Hong Kong
- Complete Library of the Four Treasuries — Huang Aiping, Academy of Chinese Studies, Hong Kong
- The Complete Library in Four Sections (Siku Quanshu) — World Digital Library, Library of Congress
- Imitation of imperial gardens in Qing Dynasty: research on Tianyige and the imitation and restoration of the seven libraries of Siku quanshu — Yu, Liu and Yang, Journal of Asian Architecture and Building Engineering, 2022
- The Emperor’s Four Treasuries: Scholars and the State in the Late Ch’ien-lung Era — R. Kent Guy, Council on East Asian Studies, Harvard University, 1987
- A Study on the Bibliographical Description of the Siku-quanshu Jianming-mulu — Lee Sang Yong, Journal of the Institute of Bibliography, 2013
- Reflections on Uses of the Electronic Siku quanshu — Ronald Egan, Chinese Literature: Essays, Articles, Reviews, volume 23, 2001
- Siku quanshu, electronic database of the Wenyuange edition — China Studies Digital Mapping Project, Digital Heritage Publishing and the Chinese University of Hong Kong
- Original volumes of ‘Siku Quanshu’ unveiled in Lanzhou — China News Service, 2016
- Literary inquisition — Wikipedia
- Destruction of Chinese Books in the Peking Siege of 1900 — Donald G. Davis Jr. and Cheng Huanwen, 62nd IFLA General Conference, Beijing, 1996
