The Dunhuang Library Cave: Manuscripts on the Move

The date is near the end. A reader who starts with the illustration at the opening of the Diamond Sutra must travel past the printed columns before reaching the words that name the occasion on which this particular copy was made. The British Library identifies the scroll as Or.8210/P.2 and dates it to 868. Its closing inscription records Wang Jie commissioning the work for distribution on behalf of his parents. In a room that held many kinds of written objects, this scroll offers an unusual foothold. Its makers left a statement that can be read alongside the thing they made.

A date can anchor a story too firmly. The year 868 belongs to the production of the printed copy. It cannot tell us when the scroll entered the small chamber near Dunhuang where it was later found. The commission tells us something about the copy's intended purpose. It does not give us the names of everyone who read it, held it, wrapped it or helped put it away. The strongest piece of evidence is often the one that makes the missing evidence most apparent.

The scroll also carries the marks of more than one method of making a book. It is paper in a format that unrolls. Its image and text were printed from woodblocks. A person can recognise the work by its words, another copy by comparing its text, and this particular object by its paper, joins, inscription and modern catalogue number. These ways of recognising it serve different ends. The text can travel independently of a copy. The physical copy carries evidence of the decisions that made it usable.

Begin here, with a sheet that has an address in a modern catalogue and a date in its own text. Then follow the gaps between those points. A cave at the Mogao site once held this scroll beside other things, and modern institutions now hold pieces of what was there. Between the cave and the catalogue lies a history of keeping, choosing and carrying. Each action left some evidence and erased the visibility of something else.

Mogao is a complex of caves near Dunhuang in what is now Gansu province, China. The chamber known as Cave 17 opens off the approach to a larger cave. Its small size matters when we picture the discovery in 1900. The objects were stored in a confined space. Anyone reaching for one item had to contend with its neighbours and the limits of the room. To call it the Library Cave is convenient. The name can also invite us to imagine shelves, an accession register and a librarian's system for finding a book. We have to earn each of those details from evidence.

Wang Yuanlu, a Daoist priest who cared for the site, was central to the reopening and subsequent handling of the chamber. His portrait, made by Aurel Stein and catalogued by the International Dunhuang Programme, is evidence of a later encounter. It gives the viewer a person and a setting. It cannot give us the scene as it appeared in 1900. The widely reproduced image of Paul Pelliot examining papers amid piles in the cave similarly records a later moment. By the time that photograph was made, earlier handling had already altered the contents. It is worth looking for the date in a photograph's caption before treating its details as a record of discovery.

A room has relationships that photographs and inventories record unevenly. A wrapper could mark the boundaries of a group of papers. A small fragment beneath a roll might have belonged to it, or it might simply have landed there. The strength of the inference depends on whether someone recorded the connection at the time and whether the pieces themselves support it. Once the items are placed in separate boxes, later readers can still investigate the question, but the position of one thing beside another is harder to recover.

Even the sealing of the chamber raises questions at different scales. One question is when access was closed. Another is why the contents came together in the first place. A date written on one document can establish that the document was made by a particular time, subject to how the inscription is read. It cannot supply the reason someone stored a much older item beside it. Think of the cave as an event in the lives of many objects. They shared a location for a period. They could have reached that location by many routes.

A scroll asks a reader for space and two hands. It shows a stretch of writing at a time, and revisiting a passage may mean rolling back through several lengths of paper. A concertina manuscript opens along alternating folds. A reader can expose one panel, a few panels or much of the length. The International Dunhuang Programme presents folded books from its collections beside scrolls. Viewed together, the objects make it possible to ask how a text was handled before we ask what its words meant.

Folds also have an afterlife. Paper repeatedly opened in the same place may weaken there. A damaged fold can leave a fragment with writing on it but very little sense of its former neighbours. Imagine finding two such panels in different folders. A line of text might suggest they once followed one another. If the edges and fold direction also agree, the case becomes stronger. A reader who understands the language can offer one part of the argument, while a conservator looking at the paper can offer another. A catalogue is valuable when it preserves the trail from suggestion to evidence.

Wrappers and ties are similarly easy to overlook. They may carry little writing compared with a fine scripture. Yet they answer a question about how several pieces became a portable group. A string holds things together without requiring them to be written by the same person or to discuss the same subject. A label might describe a bundle at one point in its life. It need not describe every object currently filed under that name. The physical arrangement and the written description are separate witnesses.

Cave 17 also contained paintings and textiles. A survey framed solely as a history of books would miss the way different materials shared space. A painting could be rolled or folded. A textile could protect an object, be an object of devotion in its own right, or survive as part of a more complicated assemblage. Keeping those possibilities distinct helps us see why a list of texts is only one way to describe the room.

Stein obtained material from Wang in 1907. Pelliot followed in 1908 and obtained another substantial group. These episodes are sometimes compressed into a line of explorers' names, but the acts of choosing differed. The choices depended on what was available, what a visitor could recognise, what he considered valuable, and what an expedition could carry. The surviving groups therefore tell us something about the interests and practical circumstances of those visitors as well as about the objects themselves.

Wang appears in these stories in more than one role. He was a person caring for a large site and a participant in transfers of its contents. A photograph and a visitor's account can illuminate those roles from particular positions. Neither gives direct access to everything Wang thought. If a modern reader wants to evaluate a transaction, the first useful questions are concrete. Who controlled access at the moment? What was recorded as having changed hands? Where did the items go? Which parts of the account come from the person acquiring them?

Language mattered at the point of selection. Someone who could quickly read a page could notice names, dates or an unusual text. Another person might judge from the material, the format or the apparent completeness of a roll. Neither method produces a neutral sample. The collection that results reflects the way it was assembled. A manuscript overlooked during one visit might later become important because another reader recognises a familiar hand or connects it with a record elsewhere.

The British Library's Stein records and the BnF's Pelliot holdings give us different routes into the surviving material. They also show why the modern address of an object is only one stop in the history. A manuscript can be associated with a findspot, an expedition and an institution, and those labels each answer a different question. The institution holding it today can provide its current shelfmark. That shelfmark is a practical way to locate the object, and a place from which to ask about earlier custody.

Consider a reader making a small notebook of Cave 17 objects. On one page, they write the Diamond Sutra's modern shelfmark. Below it they note the object's format, printed scroll, and its stated date. In the margin they record the source of each detail. Another entry describes a folded manuscript. Its shape makes comparison useful, but the reader leaves its date blank until a source establishes one. A blank field is better evidence of careful work than a date borrowed from a neighbouring object.

Now suppose two descriptions appear to name the same scripture. One record refers to a complete roll. The other describes a few surviving panels. These could represent two copies of a widely circulated text. Before proposing that the panels broke away from the roll, compare the description of the script, the layout, the surviving dimensions and the expected reading sequence. If the information available cannot settle the matter, keep the records separate and explain the reason for the comparison. A useful notebook allows a future reader to disagree with it without starting over.

Names can introduce similar problems. A repository may present a title in a translated form, in Chinese characters or in a romanisation chosen when a record was first made. A changed spelling may conceal a match. A familiar spelling may suggest a match where the physical descriptions disagree. The best safeguard is to keep both the title and the stable identifier wherever possible, and to link to the record rather than rely on a search result that may show a shortened label.

This method also keeps provenance from turning into a single vague word. Write down the claim you mean. Is it the place where a particular object was recovered? The earlier place where the paper or book may have been made? The acquisition through which it entered a collection? Distinguishing those claims exposes the evidence each one needs. It also lets you see a genuinely long journey when there is enough evidence to draw one.

The International Dunhuang Programme brings together institutions with Dunhuang holdings and makes records and images easier to consult. The BnF describes its participation and makes material from its own collections accessible through its digital services. For a reader, this means a useful comparison can begin in a single afternoon without a journey between reading rooms. A photograph can also be enlarged to inspect a fold or a join that is difficult to see at the scale of an exhibition image.

A screen changes scale with a gesture. That convenience can hide the actual size of the objects. A fragment and a long scroll might each fill a tab. If you compare them, look for dimensions in the records, and notice whether the photograph shows a whole object or just one section. A line that looks continuous across two images may conceal a gap at the join. Conversely, two images catalogued separately may show opposite sides of the same piece. Images work best alongside the notes that explain what the camera has shown.

A digital gathering can restore the possibility of asking questions across collections. It cannot put the objects back in their original positions or settle the history of every transfer. Questions of custody and the work of describing a dispersed archive remain visible in the records. A good digital edition can help a reader see those questions plainly by keeping the shelfmark, repository, image and history attached to one another.

Return to the scroll at the end of this route. The inscription names the commission, the paper carries the printing, and the shelfmark makes it possible to find this object again. These are three distinct acts of keeping information with a thing. Between them lie stretches of its journey for which we have no comparable label. The work of reading the Library Cave begins when we can say which part of a journey a piece of evidence describes. It continues when we leave enough of a trail for someone else to follow the object farther.

The British Library's International Dunhuang Programme provides collection descriptions for [Cave 17 objects](https://idp.bl.uk/discover/learning/dunhuang/collection-items/cave-17-the-library-cave/) and the [Stein holdings](https://idp.bl.uk/discover/collection-categories/stein-collection/), including the printed Diamond Sutra and its identifier. Its [account of the reopening](https://idp.bl.uk/discover/learning/dunhuang/dunhuang-articles/cave-17-the-library-cave/discovering-cave-17/) explains why the photographed arrangement cannot be treated as the original one.

For Pelliot's visit and the later institutional records, see the BnF's [Pelliot and Dunhuang](https://heritage.bnf.fr/france-chine/paul-pelliot-1878-1945-et-dunhuang), [holdings guide](https://www.bnf.fr/fr/manuscrits-du-monde-asie-oceanie) and [programme description](https://www.bnf.fr/en/international-dunhuang-programme). The comparisons with hypothetical fragments in this book are reading exercises. The surviving objects and photographs discussed in the earlier chapters are identified through the collection sources above. Sources checked 4 November 2026.