- Home/
- Literati Arts/
- Chinese Poetry and Literature: Forms, Contexts and Reading Paths/
- Chinese Leishu (Classified Books): How Knowledge Was Excerpted and Ordered
Chinese Leishu (Classified Books): How Knowledge Was Excerpted and Ordered
中国类书导读:知识怎样被摘录与分类

- What a leishu is—and why “encyclopedia” is only approximate
- Classification is an intellectual technology
- Yiwen leiju : a reference aid for composition
- Song imperial projects had different purposes
- Excerpts can preserve—and distort
- Yongle dadian : copying at unprecedented scale
- Gujin tushu jicheng : classification in print
- Keep neighbouring genres distinct
- An eight-step method for reading a leishu
What a leishu is—and why “encyclopedia” is only approximate
“Books arranged by categories”
Leishu 類書 means, literally, “category books.” The convenient translation “encyclopedia” can suggest newly written explanatory articles, but many premodern leishu assemble quotations or longer passages from earlier works beneath a hierarchy of headings. Harriet Zurndorfer’s history of the Chinese encyclopaedia emphasizes this practice of collecting, selecting and protecting excerpts. Read a leishu first as an organized field of citations, not as a modern summary of settled facts.
Selection creates a new work
Compilers decide which books count, where a passage belongs, how much context to copy and what order the entries should follow. Those decisions make an argument even when no continuous essay appears. The result is neither a neutral container nor merely a heap of fragments. It is a new reference structure built from older texts, shaped by a court, school, market, editorial team or particular task.
Classification is an intellectual technology
The table of categories is a map
Begin with the highest divisions, then follow one topic into its subcategories. A sequence from heaven and earth to rulership, offices, things, plants or animals may make hierarchy visible; another work may narrow itself to literary composition or political precedent. Xiaofei Tian’s study of literary learning through encyclopedias and epitomes shows how compilation helped define what a cultivated writer should be able to retrieve and reuse.
A heading changes the quotation
The same line placed under “wind,” “omens” or “poetic forms” does different work. Copy the category path before copying the passage. Then record the cited source, the compiler’s connective words and the neighboring extracts. Without those layers, a reader can mistake the source author’s claim for the classifier’s judgement—or overlook the way juxtaposition turns several old passages into a fresh pattern.
Yiwen leiju: a reference aid for composition
Earlier discussion and literary models meet
The Tang court commissioned the Yiwen leiju 藝文類聚 in the early seventh century, with Ouyang Xun among its leading compilers. The bibliographical guide to the work describes one of the earliest extant leishu: excerpts of discussion and polished phrases on many subjects, conceived as an aid to composition. Its design lets a writer locate inherited knowledge and literary examples together.
Read the citation label, not just the elegant phrase
Open the transmitted Yiwen leiju text and choose one heading. Separate the explanatory extracts from the poems, rhapsodies or inscriptions that follow. A memorable line is not self-authenticating: its label may point to a work whose received text differs or no longer survives. Compare the quotation with another witness before treating it as the author’s exact wording.
Song imperial projects had different purposes
Taiping yulan and Cefu yuangui are not interchangeable
The Northern Song court sponsored several enormous compilations. Johannes Kurz’s study of the Taiping yulan and Cefu yuangui places them within post-unification cultural projects. The former ranges widely across classified knowledge; the latter concentrates more strongly on political and historical precedents. Shared scale and patronage do not erase differences in source base, organization or intended use.
Compilation is also administration
Teams had to obtain books, divide labour, excerpt, check, copy and present thousands of units. Imperial sponsorship could mobilize collections and personnel unavailable to one reader, while also framing knowledge for rulership. The digitized Taiping yulan is useful for exploration, but its site warns that OCR text was generated automatically. Search results are an index to inspect, not a replacement for the scanned page.

Excerpts can preserve—and distort
A lost work may survive inside later categories
Because compilers named and copied earlier books, leishu preserve many passages absent from surviving standalone texts. That makes them indispensable for reconstruction and collation. Yet survival is selective: one sentence may have been chosen because it suited a heading or a composition exercise. A recovered fragment proves that wording circulated in the compilation’s source chain; it does not automatically restore the lost book’s full argument or original order.
Attribution is evidence to test
Titles change, homonymous works are confused, quotations are shortened and copying introduces variants. Compare several leishu, catalogues, histories and extant editions where possible. Our guide to Chinese anthologies and transmission offers a parallel warning: preservation and reframing happen together. Always distinguish the excerpt’s wording, its label and your reconstruction.
Yongle dadian: copying at unprecedented scale
A vast handwritten reference work
Compiled between 1403 and 1408 for the Yongle emperor, the Yongle dadian collection at the Library of Congress represents an immense attempt to gather knowledge in book form. Its entries could preserve long stretches or whole texts rather than only brief excerpts. That variation is why “category book” is safer than imagining one fixed recipe for every leishu.
Digital access does not recreate the lost whole
The Library of Congress has digitized its forty-one volumes, but they are a small, discontinuous survival from a once much larger manuscript compilation. Browse volume, heading and physical page together. A searchable image or isolated leaf can reveal organization and scribal practice, while it cannot stand for every subject once included. Loss itself belongs to the history of the work.
Gujin tushu jicheng: classification in print
A Qing compilation built for extensive navigation
Compiled under Chen Menglei and others and printed with movable copper type in 1726–28, the Gujin tushu jicheng described by Princeton’s East Asian Library gathers excerpts under thousands of subsections from an eighteenth-century standpoint. Its printed scale, divisions and illustrations made classification materially visible in a way different from the unique handwritten Yongle dadian.
Print stabilizes a route, not a final truth
Indexes, repeated page design and movable type help readers move through a huge work, but every route still reflects editorial selection. Compare the printed category with the earlier source and note whether an illustration, verse, biography or technical passage has been detached from its original neighbours. The guides to thread-bound books and prefaces and afterwords show how material form and paratext direct reading.
Keep neighbouring genres distinct
Anthology, notebook, catalogue and leishu
An anthology selects works for a literary collection; a biji collection gathers authored notes and reports; a catalogue describes books and their arrangement; a leishu reorganizes cited material for retrieval by category. Real books may cross these boundaries, but naming the dominant operation prevents “miscellaneous old book” from becoming a meaningless label.
Use the genre that answers the question
For literary judgement, compare shihua; for expansive description, fu; for local institutional memory, gazetteers. A leishu is strongest when the question concerns categories, excerpt chains, reusable precedents or the survival of source wording—not when it is asked to supply an unmediated picture of daily life.
An eight-step method for reading a leishu
Trace the excerpt back
First identify edition and page. Second copy the full category path. Third distinguish compiler text from quotation. Fourth record the attributed source exactly. Fifth compare another witness. Sixth restore the source passage’s neighbours. Seventh note variants, omissions and added headings. Eighth state what the excerpt can support—and what remains an inference.
Read the architecture as well as the fragment
Move outward from one excerpt to its section, neighbouring categories, preface, production history and later reuse. Continue through the Chinese poetry and literature route and the Wenxin diaolong guide to compare classification with literary theory. A responsible reading values what leishu preserve while keeping selection, hierarchy and textual uncertainty visible.