Ghost Entries: Lost Books Visible Only Through the Catalog

The Paradox of the Review That Outlived Its Book

The Siku Quanshu Zongmu Tiyao (四庫全書總目提要) contains roughly 10,000 entries. About 3,460 of these are texts that were selected for full transcription into the imperial library. The remaining 6,700 or so were relegated to the cunmu (存目)—“preserved titles”—meaning the compilers reviewed them but judged them unworthy of copying. For the included texts, seven handwritten copies were made and distributed to imperial libraries across the empire, giving them a reasonable chance of survival. For the cunmu texts, no copies were made. They were returned to their owners or their provincial libraries and left to whatever fate awaited them.

Fate, as it turned out, was not always kind. Wars, floods, fires, the Taiping Rebellion (1850–1864), the fall of the Qing in 1912, the upheavals of the twentieth century—the accumulated destruction of two and a half centuries eliminated countless texts that had been circulating freely when the Siku compilers handled them. In some cases, the only evidence that a text ever existed is the catalog entry itself.

These are the ghost entries: reviews of books that are no longer there. The critics' assessment is the last trace of the work they assessed.

What a Ghost Entry Preserves

A typical Siku catalog entry, even a brief one, contains a surprising amount of information. The standard format includes the title, the author's name and dynasty, the number of juan (fascicles), and then a critical review. The review typically summarizes the book's contents, identifies its main arguments, assesses its quality, notes any textual problems, and situates it within the broader tradition of its field.

For a surviving text, this is useful context. For a lost text, it is everything. The review becomes our only window into what the book contained. We know its scope because the compilers described it. We know its arguments because they summarized them. We know its quality because they judged it. We even know, in many cases, which other texts it drew on and which traditions it belonged to, because the compilers' critical method involved situating every work within its intellectual genealogy.

A ghost entry is not the book. It is a scholar's memory of the book, compressed into a paragraph. But it is enough to tell us that the book existed, what it was about, and what someone who actually read it thought of it. In the archaeology of lost knowledge, that is substantial evidence.

How Books Disappear

The mechanisms of textual loss in China are well documented. Fire was the most common destroyer—woodblock-printed books and handwritten manuscripts are combustible, and Chinese cities burned with depressing regularity. Warfare was second: the destruction wrought by the Taiping Rebellion alone eliminated entire provincial library collections that had survived intact for centuries. Political upheaval was third: the fall of a dynasty typically disrupted the institutional networks that preserved and circulated texts.

But there is a fourth mechanism specific to the Siku project itself: the Literary Inquisition (文字獄). The same project that cataloged tens of thousands of texts also identified and ordered the destruction of texts deemed politically dangerous—primarily those containing anti-Manchu sentiment. In some cases, the compilers reviewed a text, wrote a catalog entry, and then the text was confiscated and burned. The review survives. The book does not.

This creates a peculiar category of ghost entry: the text that was killed by the same institution that preserved its memory. The catalog becomes both the executioner and the obituary. The compilers recorded what the text contained, assessed its literary or intellectual value, and then—because it contained a passage that offended the Manchu court—consigned it to destruction. The review is the scar left by the cut.

What We Can Infer

Modern scholars have developed sophisticated techniques for extracting information from ghost entries. The compilers' critical methods were consistent enough that their reviews follow predictable patterns, and those patterns can be reverse-engineered to reconstruct aspects of the lost text.

When the compilers say a text “follows the methods of” a known school or tradition, we can infer its general intellectual framework. When they note that it “draws heavily on” a specific earlier work, we know something about its sources. When they quote a passage—even a single sentence—we have a direct fragment of the lost original. When they describe its structure (how many chapters, how the argument is organized), we can reconstruct its architecture. When they compare it to surviving texts, we can triangulate its position within the intellectual landscape.

The ghost entries are, in effect, the compilers' field notes on a landscape that has since been partially destroyed. The landscape is gone, but the notes tell us what was there. And because the compilers were very good at their job—precise, systematic, evaluative—the notes are often remarkably informative.

The Accidental Archive

The compilers did not set out to create an archive of lost books. They were writing reviews—practical documents intended to guide future scholars and librarians. The fact that their reviews became, in some cases, the only surviving record of the texts they reviewed is an accident of history. The catalog was designed to describe an existing library. It ended up preserving the memory of a library that no longer exists.

This accidental preservation function operates across all four divisions of the catalog. Lost commentaries on the Classics, lost historical compilations, lost philosophical treatises, lost literary collections—all survive as ghost entries. The divination section (術數類) alone preserves descriptions of dozens of fortune-telling manuals and geomancy texts that have since vanished. The military section preserves descriptions of tactical manuals that were obsolete by the nineteenth century and are now lost. The medical section preserves descriptions of pharmacological texts whose remedies are preserved only in the compilers' summaries.

In each case, the review is more durable than the text it reviews. The seven copies of the Siku Quanshu itself suffered their own losses—the Yangzhou and Zhenjiang copies were destroyed during the Taiping Rebellion, and others were damaged—but the Zongmu Tiyao, being a more compact and widely reproduced work, survived in multiple editions. The catalog outlived both the texts it cataloged and several of the library copies it was meant to accompany.

Reconstructing the Lost

The scholarly use of ghost entries goes beyond simple acknowledgment that a text once existed. In the best cases, the catalog entry provides enough information to reconstruct the lost text's contribution to its field. If the compilers describe a medical text's approach to a specific disease, and that approach differs from what survives in other sources, the ghost entry preserves a data point about the diversity of pre-modern Chinese medical thought that would otherwise be invisible. If they describe a philosophical text's argument against a specific position, the ghost entry preserves evidence of a debate whose other side is now silent.

Some lost texts can be partially recovered through a technique called jiyishu (輯佚書)—gathering fragments. A text that is lost as a standalone work may survive in quotations embedded in other texts. The Siku compilers themselves practiced this method, noting when a text they were reviewing quoted extensively from an earlier work. Modern scholars use the ghost entries as a starting point: the catalog tells them what to look for, and they then search surviving texts for quotations that match the compilers' description. In this way, the ghost entry serves as a map to fragments scattered across the surviving corpus.

The Weight of Absence

To read the ghost entries is to feel the weight of what is missing. The Chinese textual tradition was once vastly larger than what survives. The Han dynasty bibliography (漢書藝文志) lists thousands of titles; most are now lost. The Sui dynasty bibliography (隋書經籍志) records thousands more; the majority are gone. Each successive dynasty compiled its own bibliography, and each records a tradition that was already partly ruined. By the time the Siku compilers did their work in the 1770s, they were cataloging what remained after two thousand years of attrition.

And then more was lost after them. The ghost entries are the most recent layer of absence—texts that survived long enough to be read and reviewed by Qing scholars but did not survive into the modern era. They are the closest ghosts, the most recently departed. We have more information about them than about any previous layer of lost texts, because the Siku compilers were more thorough and more systematic than any previous generation of bibliographers. But they are still gone.

The catalog preserves their shadows. A paragraph each. A title, an author, a dynasty, a judgment. The ghost entries are the Siku project's unintended memorial to the fragility of the very tradition it was trying to preserve.

References