To Collect and to Order: the Siku quanshu 四庫全書 and its Organization

Collecting and hoarding are distinguished by order. An agglomeration of objects is defined by chaos while a collection comes into being through its organization. The largest collection of texts undertaken in Chinese dynastic history, the Complete Writings of the Four Repositories ( Siku quanshu 四庫全書 ), is the high point of late imperial compilation projects ( congshu 叢書 ). While much scholarship has been devoted to explaining the criteria of inclusion, the question of order remains largely unexplored. In this article, I investigate the link between the collection of knowledge and its organization in the high Qing. Specifically, I explore the poetic understanding of knowledge, the intellectual, non-political purposes behind the collection and its fundamental principle of order. I end this essay offering some remarks on the nature of the Complete Writings , high Qing scholarship, and contemporary attitudes towards classification.

of knowledge. The Complete Writings of the Four Repositories (Siku quanshu 四庫全書) towers as the largest and most complex text collection project (congshu 叢書) in Chinese dynastic history. While much scholarship has been devoted to explaining the criteria of inclusion, the question of order has remained marginal. This article explores the link between the collection of knowledge and its organization in the high Qing by working through the sociocultural and intellectual (non-political) factors that informed the order of the Complete Writings.
By any measure, the collection was enormous. Almost three and a half thousand texts were copied in full, producing a master anthology so large that it effectively functioned as a library necessitating its own physical structure. Sponsored by the Qianlong 乾隆 emperor (1736-1796), the collection aimed to commemorate the greatness of all amassed knowledge while also capturing its most essential elements.
The immensity of the collection meant that the project could only be carried out collectively through the mechanisms of a highly effective and sophisticated bureaucracy. Scholars at the highest levels of imperial institutions and beyond deliberated on the fundamental aspects of the project. Their shared notion of knowledge and its order were the precondition which allowed for the production of such a collection. This article is about the shared substratum of explicit and implicit assumptions, attitudes, and values held by the people behind the project-collectively identified as the editors, collators, or makers of the Complete Writings-which underlie its formation.
The compilation was organized on the basis of the long-standing fourfold (sibu 四部) order of knowledge which divided texts into four main divisions: Classics ( jing 經), Records (shi 史), Masters (zi 子), and Collections ( ji 集). The fourfold order was not blindly adopted and, as Section 2 shows, alternative models were considered in the collection's embryonic stages. When the fourfold order was accepted, organizational issues were reworked and the overall order was refined by introducing a three-tier partition: the four main divisions (bu 部) were comprised of a total of forty-four sections (lei 類) made up of seventy groups (shu 屬) (Figure 1).
Much scholarship has been devoted to the Complete Writings. Its scholarly achievements and sustained cultural significance-it is on permanent display at the National Palace Museum in Taipei-have made it an object of study worthy of its own scholarly field (siku xue 四库學).1 Nonetheless, its classification remains poorly understood for two reasons. First, most scholarship construes classification decisions-indeed all decisions-through a political mired by "mutually conflicting [categorizations and] contradictions" (扞格和 矛盾), lamenting its "relative disorder and lack of a fixed principle" (相當紊亂, 沒有一定規格).3 These two biases have severely hampered our understanding of the Complete Writings. Sections 1 and 3 show how different cultural and cognitive conceptions of classification yield substantially different models of order. Section 2 demonstrates how appeals to political explanations cannot give a convincing account of the collection's organization.  This article approaches the question of textual collection and knowledge organization 'from the ground up' and 'on their own terms' . Joachim Kurtz has advocated for a "de-modernised" Sinology that aims to understand the implicit and explicit criteria of validity, veracity, credibility, coherence, relevance, [and] applicability […] in these fields [education, law, canonical studies, historiography] […] and record and define [their] terms, or metalanguages.4 He equates such a method with a need "to reverse the conventional perspective and try to reconstitute concrete modes of knowledge production and their underlying rules 'from the ground up.'"5 Similarly, Benjamin Elman has argued for the necessity of exploring "Chinese interests […] as they articulated and practiced them on their own terms rather than speculate why they did not accomplish what the Europeans did."6 This requires us to "acknowledge that as yet we do not have appropriate categories of learning that resemble premodern Chinese frames" of knowledge and that we will have "to extend our own understanding and make room for […] pre-modern Chinese frames."7 Investigating the order of the Complete Writings sheds light on the collection, the society which produced it, as well as our own assumptions about classification. Late imperial China was a knowledge society par excellence and examining the high Qing through the question of knowledge organization yields new questions about its intellectual and social history. Following Peter Burke's call to de-parochialize our understanding of knowledge societies, a non-teleological alternative to a singular and progress-bound conception of Knowledge is vital.8 Appreciating how knowledge in different contexts has been conceptualized, broken down, held apart, and brought together is essential in providing such an alternative.
Knowledge comes in various forms but it always functions as a nexus between world and society. How we perceive and classify natural objects determines how we order society and behave in it. Conversely, sociocultural structures frame how we perceive the world. Knowledge and its organization play a mediating, dynamic role between our conceptions of the world and our actions within society.
This article follows the tripartite structure of world, knowledge, and society. In Section 1, I show how the antecedent poetic understanding of knowledge as water inflects the order of the Complete Writings. I provide an account of its impact on classification and briefly sketch its historical origins and development. By way of comparison to the western notion of knowledge as tree, I tease out the properties and cultural specificity of each system. In Section 2, I explore the intellectual, non-political purposes behind the collection because any bibliographic arrangement fulfils a certain set of purposes which determine its structure. I provide a short overview of past bibliographic classifications and their impact on the editors' scholarly objectives and show how the order of the Complete Writings was envisioned as an advising guide to scholarship. In Section 3, I identify the ascribed socio-scholarly function of texts as the fundamental principle of order and outline the different ways in which it is expressed. I conclude by showing how the high Qing conception of order differs from Aristotelian and Wittgensteinian notions of classification. I conclude with some remarks on the nature of the Complete Writings, high Qing scholarship, and contemporary attitudes towards classification. I explore these points by drawing from the compilers' annotations to bibliographic sections; historical documents recording the formation of the compilation; relevant philosophical and historical texts; as well as the rich secondary literature on the Complete Writings.

Poetics: Knowledge as Water
The Complete Writings has been extensively studied and yet it is remarkable how some of its most fundamental aspects remain unexplored. The names of the halls housing the collection are among these blind spots. These are: The Literary Abyss Library (Wenyuan ge 文淵閣) The Literary Source Library (Wenyuan ge 文源閣) The Literary Ford Library (Wenjin ge 文津閣) The Literary Upstream Library (Wensu ge 文溯閣) The first graph wen 文-meaning 'literature' but also 'erudition' , 'learning' , 'demeanour' , and 'knowledge'-is followed by a graph denoting a water-related feature. The final graph ge 閣 denotes the physical edifice of a library hall.
Since the first and third graphs are the same and the middle graphs are all tied to water, the names strike an elegant symmetry, hard to render in English. Contemporary scholarship has treated the names as an expression of superstition wrapped in literary language. Water was chosen for its apotropaic function to protect books from fire, their perennial foe, and different water elements exhibited the makers' literary acuity.9 But, is this all that there is to the names or did the compilers have more meaningful reasons to choose waterrelated terms?
The 1980s marked a revolution in cognitive linguistics. Lakoff and Johnson's seminal book Metaphors We Live By offered a strikingly different conception of metaphor.10 For them, metaphor is not merely an aspect of language but rather an aspect of thought, and abstract notions such as time, reason, virtue, argument, and emotions are metaphorically conceived in terms of concrete objects. "Don't waste my time", for example, reflects how time is conceived as a resource. "She defended her argument", "He attacked her position", and "She won the debate", reveal how argument is structured in terms of war. Metaphors We Live By generated a flurry of scholarship that collected further evidence, refined the theory, and expanded its methodology.11 Much scholarship has been geared towards showing how conceptual metaphors are found in all languages and societies and a small yet increasing literature exists for pre-modern Chinese texts.12 While the notion of conceptual metaphor as a culturally neutral, universal mechanism of cognition is problematic, it remains useful in its suggestion that meaning is articulated on the basis of imagistic, poetic, and pre-reflective conceptions of concrete objects.13 9 Liu 1997: 125-134 presents this view in great detail. Liu Qiang 劉薔 1998 explains that the four library names are based on a widespread belief in the apotropaic function of water. For more on the library halls, see Wang Yi 王穎 2003;Wang Zhiwang 王智汪 2008;Huang Aiping 2010;Tong Zhenglun 童正倫 2012. 10 Lakoff and Johnson 2003. 11 Developments are fast-paced and conceptual metaphor theory has evolved and become more sophisticated over the years, see Lakoff 1987;Ortony 1993;Faucconnier 1997;Faucconnier and Turner 2003;Trim 2007;Glucksberg 2008;Steen 2011;Kövecses 2013;Gibbs 2015. 12 Slingerland 2003, 2004a, 2004b. 13 Gandolfo 2019 In western societies, knowledge has historically been conceptualized as a tree.14 The Biblical Tree of Knowledge and the various manifestations of this notion have been, according to several cultural theorists, foundational to 'western' civilisation.15 The idea that knowledge is organized in branches is so engrained that the expression does not even strike us as metaphorical. In the frontispiece of the 1780 edition of the Enlightenment Encyclopédie, the Tree of Knowledge is rendered in all its glory and conceptual specificity ( Figure 2). The large branches, Philosophy, History, and Poetry, lie at the lowest end and carry all other smaller branches, creating a hierarchical nesting order. The structure of trees is imposed onto knowledge with binary divisions resulting in the placement of different branches of learning farther apart from each other. Knowledge as a tree is not a helpful illustration but rather, as evinced by its schematized form (Figure 3), fundamental in how the order of knowledge is conceptualized.16 Conceptions of knowledge in pre-modern China were significantly different. The term zhi 知 is the standard translation for knowledge but its semantic range covers 'realization' , 'perception' , and 'wisdom' . In pre-modern Chinese, the distinction between knowledge and wisdom was blurred and the nearhomophonous graph zhi 智, reserved in modern Chinese for wisdom, was used interchangeably with zhi 知. Barry Allen stresses this link by holding that knowledge in the Chinese setting "poses questions about point and value".17 Ancient Chinese thinkers did not aim to deploy formal definitions of 'knowledge' or elucidate its relationship to the senses, mind, or reality. Instead, as Allen points out, Chinese texts pondered on "what makes knowledge wise and worth pursuing? What is the relationship to other values-good government, say, or ritual, or war?"18 Allen treats the relationship between zhi and sheng 聖, sagacity, as foundational in classical China suggesting that penetrating the subtle-seeing a lot in little things [and] discerning the concordant and contrary-knowing the resonance among things, and how to amplify or dampen emerging tendencies […] is the ecumenical 14 Gontier 2011;Lima 2015. 15  understanding of sage knowledge in the Chinese tradition; it is not strictly universal-there are exceptions-but it is the leading idea.19 In the quest for sagacious knowledge, wen 文, erudition and demeanour, plays a crucial role. Even though different philosophical traditions have distinct notions of sagacity, wen functions as a minimal common denominator. As previously noted, the term has a broad semantic field and for this reason can be regarded as the manifestation of zhi-namely, that which has been apprehended. In the form of text, wen is the material manifestation of zhi; in the form of demeanour, wen is its physical embodiment. Wen captures zhi and transmits it through language and action. The conceptual links between zhi, sheng, and wen are not accidental. The terms are related because they are based on the same conceptual imagery. In The Way of Water and Sprouts of Virtue, Sarah Allan shows that dao 道-the fundamental pattern of change in the world-is metaphorically grounded on water.20 Texts across traditions refer to the flow of dao, its spontaneous twists and turns, its abysmal and primordial quality, its pure, all-encompassing nature, and its ability to create and sustain life. Knowing dao and putting it into practice is structured on water and time and again a wide range of texts refer to the ocean-like wisdom of the sages, their capacity to extract ethical insights from water and structure their actions on the free and easy flow of a river. Dao is conceived in water terms and so too are the knowledge, sagacity, erudition and demeanour that come with it.
The library records of the Complete Writings build on the conception of knowledge as water. The Record of the Literary Ford Library (Wenjinge ji 文津閣記) states that "water is used as a metaphor to understand knowledge" (以水喩文) and explains the structural similarities between the four main divisions and the four water elements of the library names.21 The meaning of the names, the record notes, is established on water, locating their signification not on wen, the prima facie conceptually rich term, but rather on the seemingly innocuous 'water ' .22 The primary similarity between water and knowledge lies in the order of the four main divisions. The Classics are the source (yuan 源) from which all other streams of knowledge flow out.23 With the Classics as the source of all knowledge, the water structure confers a clear hierarchy to the organization. The Classics are at the top-conceptually and in the order of the collectionand the emanating streams flow out in a ranked sequence: the Records are the main current (liu 流), the Masters are the tributaries (zhi 支), and the Collections are the smaller brooks and runnels (pai 派).24 Each division is more important than the one that follows it. The source itself is understood as the abyss (yuan 淵)-a reference to dao-guaranteeing the ceaseless flow of all other streams.25 The implication is that Classics are the direct manifestation of dao and that the farther down in the order, the farther away one is from dao. Placing the Classics as the source of knowledge was a way to enshrine their intellectual pre-eminence. But it was also a recognition of their significance in the historical development of learning. It was an act that simultaneously reflected as well as consolidated their significance in the generation of knowledge as it had flowed over the millennia.
Having established this premise, the rationale linking water to the order of knowledge takes an interesting turn, revealing a deep concern for tracing knowledge back to its source.26 The way upstream is not simple to find: one first needs to know (zhi 知) the location of the fords-the conceptual links which allow one to cross from one side to the other and move up to the source. There are many paths, and, the editors caution, without knowledge of the fords one will inevitably be lost.27 But while the movement upstream was essential in the conception of knowledge as water, so was the natural downward flow of a river towards the sea. Representing the collective wisdom of the sages, the sea symbolized the purpose of the acquisition of knowledge.28 One did not learn merely to become more knowledgeable but rather to become wiser, to reach closer to the ethical and political ideals set by the sages of antiquity.29 Entering in the right stream of knowledge-and that was guaranteed by the editors' careful selection of texts-ensured the natural flow towards sagacious knowledge. What this conception of knowledge meant in practice was that as long as one kept 24 Ibid. 25 Zuanxiu, 2.2723. 26 Ibid. 27 Ibid. 28 Most notably in Mengzi 孟子: "Therefore, one who has contemplated the sea, finds it difficult to think of other waters, and one who has wandered within the gates of the sages, finds it difficult to think of the words of others" (故觀於海者難為水遊於聖人之門者 難為言). In Siku quanshu: Mengzi zhushu, (13 xia 下), (henceforth SKQS; in parenthesis is the original juan 卷 number). 29 Allen 2015: 3. In the context of the Complete Writings, see Guy 1987: 57. oneself in the streams of knowledge-which, make no mistake, required endless effort-one would naturally come closer to the sages' ideal. Sagacious knowledge was not something one willed into action but rather something which emanated by its own accord.
To drive the point home that knowledge is structurally similar to water, the rationale concludes with a rhetorical question: "with the essence-and-function of water being such, could the essence-and-function of knowledge (wen) alone not be like this" (水之體用如是文之體用顧獨不如是也)?30 Far from being apotropaic, aesthetic, or trivial, the library names embody an essential element of the collection.
The compilers were not the first to use the language of water to conceptualize the order of knowledge. The Annals of the Former Han (Qianhan ji 前漢紀) note that Liu Xiang 劉向-composer of the Seven Epitomes (Qilüe 七略), the first formal classification of knowledge-canonized and collated classical texts into nine streams ( jiuliu 九流).31 In the Writings of the Sui (Sui shu 隨書), compiled in the early seventh century, water imagery was extended to discuss the source and flow (yuanliu 源流) of a particular stream of learning. In his magnum opus, Record of Continuities (Tongzhi 通志), the Song dynasty scholar and master bibliographer, Zheng Qiao 鄭樵, argued that by examining the texts of a period one could know the source and flow of its scholarship.32 Bibliographic entries in the Catalogue of the Hall of a Thousand Acres (Qianqing tang shumu 千頃堂書目), privately compiled in the seventeenth century, employ the language of source and flow to refer to smaller and larger streams of learning such as the source and flow of the Book of Changes (易學源流)33 or the study of the classics (經術源流).34 By the late imperial period, knowledge as water had found its way into the titles of texts indicating how deeply and widely the imagery was absorbed into scholarly practice-the Qing dynasty Treatise on the Source and Flow of Medicinal Studies (Yixue yuanliu lun 醫學源流論)  What distinguishes the Complete Writings from past instantiations of the water as knowledge framework is the deliberate and explicit connection between the two.36 The compilers were not unconsciously employing a deeprooted cultural frame. By naming it and drawing attention to it, they were bringing it to the surface. Their use of the water structure was a mindful, second-order application of the concept which drew from and expounded on its inheritance. There was, of course, no necessity that required the Complete Writings to be premised on water just was there was none for the Encyclopédie to be premised on trees and, in both contexts, knowledge was conceptualized in multiple ways.37 Nonetheless, it is the case that a leading conception of the order of knowledge developed, matured, and consolidated in water and tree terms.
Despite their differences, the Complete Writings and the Encyclopédie reflect the same attitude towards classification: both assume their validity and make the case for it by putting it into practice-a position which can be described as taxonomical realism. In the Encyclopédie, 'organizing knowledge' is dividing it in ever smaller branches; in the Complete Writings, 'organizing knowledge' is ordering its streams and tracing its flow. Even though trees favour binary divisions and water allows for flexibility and continuity, they both 'perform' 36 Zuanxiu, 2.2724-5. 37 In China, alternative metaphors such as mirrors, lamps, and plants were used in part to structure conceptions of knowledge. For more on these metaphors, see Lai 1979;Oshima 1983;Allan 1997: 93-122;Liu Yi 劉藝 2004;Chen Lisheng 陳立勝 2005, 2009 what they state, implicitly yet powerfully affirming the validity of their own knowledge order.38

Purposes: the Complete Writings in History
In early January 1773, Zhu Yun 朱筠, the Anhui education commissioner, submitted to the throne a proposal for a massive bibliographic collection. The proposal was unusual because education commissioners only used the imperial memorial system to submit lists of successful examination candidates.39 Zhu Yun's proposal was particularly odd because, even though responding to an imperial edict, it was directly addressed to the Qianlong emperor.40 Despite its irregularity, the proposal was admitted and carefully reviewed because, after all, Zhu Yun was one of the most prominent scholars at the time and his proposal was significant and exciting.41 Yet Zhu Yun's memorial stands out for another reason, overlooked by the scholarship. In his proposal, Zhu Yun explicitly raised the question of the collection's organization stating that it could be arranged according to the sevenfold organization of the Qilüe or the fourfold division of the sibu.42 This seemingly innocuous option, however, has major implications for how we understand the Complete Writings and its purposes. The possibility of following different ordering systems highlights the fact that, in its embryonic stages, the Complete Writings was not tied to any one organizing system.
The suggestion to employ the Qilüe was radical for it would constitute nothing short of a sweeping rearrangement. Befitting its original Han dynasty context-or rather the goals of Han imperial scholars-the Qilüe order had been abandoned for almost fifteen hundred years.43 And arguably for good reason, since, among several issues, the Qilüe did not have a special section for historical texts but gave outsized importance to the militarist school which had long faded. Proposing to revive such an outdated system must have appeared bold as well as mind-boggling to many high Qing scholars.44 38 For more on text performativity in the Chinese context, see Meyer 2012. 39 Guy 1987 For more on Zhu Yun, see Hummel 1991: 198-199. For more on the operations of the Qianlong court, see Bartlett 1991. 42 Zuanxiu, 1.20-21. 43 For more on the Qilüe, its use, and survival, see Lee 2016. 44 Luo Mengzhen 羅孟幀 1986. Other scholars such as Zhang Xuecheng 章學誠 and Qian Daxin 錢大昕 argued for the revival of the Qilüe yet both eventually dropped or Zhu Yun's proposal was a product of Han learning (Hanxue 漢學), a scholarly movement which had effectively merged with evidential studies (kaozheng xue 考證學), after about two centuries of fermentation, experimentation, and consolidation.45 Han learning aimed to recover and reconstruct the past using-in modern diction-philological tools and draw from Han dynasty scholarship, ideas, and institutions to expunge the additions and alterations of later scholars. Han learning was diametrically opposite to the more speculative, subjective-and philosophical-understanding of classical texts which was associated with the scholars of the Song, known as Song learning (Songxue 宋學). The evolving relationship between these two approaches to the sagely past was one of the fundamental features of the Qing intellectual landscape.46 Beyond Zhu Yun's proposal to revive the Qilüe, there was a plethora of organizing systems. During the Song and Ming dynasties private and imperial collections employed orders that differed significantly from the fourfold division. In the thirteenth century, Zheng Qiao marked a first by developing a systematic theory of bibliography, its methods, and purposes.47 He also harshly criticized the sibu and Qilüe and deployed an innovative three-tiered system with twelve main divisions and over five hundred sub-and sub-sub-categories.48 During the Ming dynasty, the renowned scholar and bibliographer, Jiao Hong 焦竑, in the Bibliographic Treatise ( jingjizhi 經籍志) chapter of Records of the State (Guoshi 國史), adopted Zheng Qiao's division but also incorporated the main bibliographic development of his era which saw the displacement of the classics from their top position by the writings of emperors of the ruling dynasty.49 The bibliographic landscape of the six centuries prior to the compilation of the Complete Writings was defined by plurality and cross-pollination reflecting the vibrant and robust scholarly culture of late imperial China. The fourfold order changed their views. Naito Konan 內藤湖南 1970: 432-437 argues that Zhang Xuecheng realized that his ideas were practically unrealizable. Elman 1984: 73, 165 presents Qian Daxin's position but concludes that Qian following Zhang "admitted that Liu Xin's categories [i.e. the Qilüe] were no longer appropriate for modern cataloguing." 45 Han learning was a regional, sub-branch of evidential scholarship. In time, however, Han learning came to be used metonymically for the entirety of the new philological movement. For more on regional differences, see was a staple of late imperial scholarship but it existed in a competitive and syncretic environment whereby its eventual adoption in the Complete Writings cannot be treated as a foregone conclusion.50 And yet most scholarship has overlooked the alternatives at hand, assuming the fourfold order was chosen because it embodied orthodox practice.51 However, it is far from obvious why the fourfold division, enshrined in the Tang, would be considered more orthodox than the Qilüe of the Han. Invoking 'orthodoxy' without explaining its antecedent factors of formation is meaningless. More crucially, sociopolitical considerations did not exhaust the makers' array of concerns. A strict political reading fails to take note of its scholarly dimensions and a theoretical layer must be added to account for the adoption of the fourfold order vis-à-vis the collection's purposes. What did the makers think the Complete Writings should achieve and why was the fourfold division-and not its alternatives-deemed better at realizing these goals? To answer this, we first need to establish who the envisioned user was.
Nominally, the collection was intended for the Qianlong emperor and his successors. By association, a small group of elite imperial scholars would also presumably have access to it. However, the answer is complicated by the participation of the larger scholarly community, which points towards a broader conception of its use. Zhu Yun's memorial shows how non-imperial scholars influenced the formation of the collection from its earliest stages. The contribution of book-owning families who submitted their rare books for transcription and the eventual decision to produce three copies of the collection as a reward came to underscore the collection's public character.
The nature of collecting itself paved the way for the idea to disseminate the Complete Writings. The very basic purpose of amassing, reproducing, and publishing texts in one master anthology sanctioned the public orientation of the project from its inception. Calls for the formation of a Confucian Canon (ruzang 儒藏)-which would stand as an equal to its Buddhist and Daoist counterparts-were intensifying. In the Qianlong era, Zhou Yongnian 周永年 argued that the fundamental purposes of such a project would not only be the preservation and transmission of texts but, more importantly, the enrichment of scholarly pursuits and the improvement of imperial governance. for the collection of texts which clarified the essentials of governance or illuminated human nature and the morals of the world.53 Similarly, in his proposal, Zhu Yun argued for the need to outline the sages' vision in the effort to restore antiquity.54 By undertaking a task of such monumental proportions, the makers of the Complete Writings must have felt-indeed known-that their work would be used and commented on by later scholars, just as they had done with their predecessors. The sustained use of the Complete Writings as a reference tool to this day highlights its intellectual achievements as a piece of publicly accessible scholarship. The intergenerational-indeed inter-epochal-aspect of the collection underscores its public orientation and the fact that its intended users were scholars across space and time.
The collection was envisioned not only as public but also ecumenical. The introductory preface to the Classics reveals a conciliatory stance between Han and Song learning.55 These were not two traditions to be pitted against each other but rather, recognizing their strengths and weaknesses, to be synthesized into a balanced approach, dissipating their differences. This new attitude, the editors argued, was necessary to bring forth the public principle and expunge the biases of individual minds in order to illuminate the meaning of the classics.56 Adopting the order of the Han dynasty Qilüe would have signalled a total victory for Han learning and failed to uphold the synthetic, public, and ecumenical spirit of the collection. The fourfold order did not have particular associations with either tradition. More importantly, however, since it had been broadly used in late imperial collections, it remained one of the most recognized systems whose internal logic was arguably known to the widest crosssection of scholars.57 Beyond the issue of obsoleteness, the exclusion of the Qilüe was premised on grounds of public and ecumenical accessibility, rather than orthodoxy and legitimacy. The rejection of the more detailed orders of the Song and Ming dynasties was premised on different grounds which, however, remained within the bounds of scholarly deliberation. In the introductory Guidelines for Readers ( fanli 凡例), the editors provide a telling-and rather negative-evaluation of Jiao Hong's division. The main objection is that it has too many items, making it overloaded.58 The subdividing streams employed in sections such as philology, geography, biographies, genealogies, and the arts, among others, are too numerous and the general structure ultimately leads to confusion.59 Instead, they argue that the divisions used in the Complete Writings-which are fewer in number and less detailed-not only offer an orderly arrangement but also facilitate readers with searching through the catalogue. Compared to Jiao Hong's division, the Complete Writings, by deleting and merging different sections, had dispensed with trivial minutiae.60 The compilers were responding to what appeared to them as taxonomical hypertrophy, the debilitating explosion of categories that renders classification meaningless (for using ten categories to organize ten objects amounts to not organizing them at all).61 In their eyes, classification orders such as Jiao Hong's had too many divisions which needed to be simplified. It was for this reason, and not because of orthodoxy or legitimacy, that they were rejected.
With criticizing such ordering systems, the compilers were challenging a conception of bibliographic order which had emerged with the writings of Zheng Qiao. Zheng Qiao prized division as a form of clarity, arguing that increasing granularity made the meaning of scholarship self-evident.62 It is for this reason that he was opposed to the practice of critical introductions.63 There was no need to explain the content of any one text since the clarity of the division would provide enough information. His critique against critical fourfold division. Furthermore, private collections such as the twelfth-century Records on Reading of the Prefectural Hall ( Junzhai dushu zhi 郡齋讀書志), the Catalogue of the Hall of Retirement (Suichutang shumu 遂初堂書目), and the Annotated Bibliography of the Hall of Rectitude (Zhizhai shulu jieti 直齋書錄解題) also employed the fourfold division. In short, the fourfold organization by the time of the compilation of the Complete Writings had been neither the only, unchanging, mummified way of organizing libraries nor a lifeless, distant memory that had to be revived. 58 SKQSZM: Fanli 凡例, article 5. 59 Ibid. 60 Ibid. 61 Owen 2007: 1392-1393. 62 Zheng Qiao, Tongzhi, 2.1806. 63 Tongzhi, 2.1818 prefaces underscores his stance towards bibliographic divisions as having a primarily descriptive (and not evaluative) purpose. Extending the water conceptual metaphor, we can think of Zheng Qiao's conception of classification as a precise and exhaustive descriptive map delineating the streams of knowledge.
The editors heavily criticized Zheng Qiao. Critical prefaces were good bibliographic practice not only because they allowed readers to understand texts at a glance but also because they guided them through their strengths and weaknesses.64 By the Qianlong era, the streams of learning had become so numerous and the quantity of texts so large that scholars felt drowned in a deep ocean.65 The purpose of the Complete Writings was not to provide a detailed, informational overview of knowledge. Rather, it was to act as an advising guide helping scholars move from any one stream of learning (or text, allusion, or idea) to the source of knowledge.66 The division of the Complete Writings had all of the necessary information to reach the source-everything else was superfluous or, even worse, dangerous. The most precise map is not the easiest to use and the most detailed division-let alone one without critical remarks-is not the best at guiding one's study.
The differences between the two conceptions did not mean that precision was not part of the theoretical purposes of the Complete Writings. The compilers did value detail, which is why the collection was the first-and onlyofficial bibliography to employ a three-tiered division, implicitly paying tribute to Zheng Qiao's innovation. The makers of the Complete Writings did not have an 'ideological' , predetermined stance towards past bibliographies but rather retained a subtle and complex position willing to adopt and adapt what they considered valuable. Preservation of orthodox bibliographic practice was not a priority. A public and ecumenical order which cohered with the intellectual landscape and could act as an advising guide through the streams of knowledge was far more important. To categorize is human and yet-or rather, therefore-what categorization is remains highly contested.67 Identified by many cognitive scientists as the most basic phenomenon of cognition, categorization has been tackled by philosophers, psychologists, anthropologists, library and information scientists, and beyond.68 Two positions dominate the spectrum of views across fields: the classical 'Aristotelian' and the revisionist 'Wittgensteinian' theories.69 In the classical view, categories are groups of objects possessing the same individually necessary and jointly sufficient properties.70 Mammals, prime numbers, and bachelors are good examples of classically conceived categories since membership can be easily established through asking a series of yes/no questions (is the person male? is he an adult? has he never married?). Categories are thus clearly defined, mutually exclusive, and totally exhaustive.71 Wittgenstein, on the other hand, proposed that the meaning of a word is determined by its use but that given the breadth of uses of any given word, it is impossible to fix its meaning. His famous example on the impossibility of defining the word 'game' revealed the cracks of the classical conception of categories. Wittgenstein proposed that the only way a category could be formed is through a set of likely features of family resemblance. Just as members of a family share only a handful of traits so too do members of a category.72 Belonging in a category is a matter of degree. Dogs, rather than snakes, are a better fit to the category of pets just as couches, rather than radios, are a better fit to the category of furniture. In this model, categories are fuzzy and overlapping.
In the preface to the Miscellaneous Schools and Writers (zajia 雜家) section, the editors state that it consists of texts which cannot form a group of their own.73 They trace the origins of this problem to the fact that, with the collapse of the Zhou dynasty, hundreds of schools competed against each other, each establishing its own writings and scholarly lineage.74 The problem of miscellaneous-ness (za 雜) arose because in later times many of these teachings were not reproduced, resulting in breaks in their transmission and the impossibility of recording them in a simple and unequivocal way.75 The comprehensiveness of the group is not attributed to heterogeneity of subject matter-in fact, content is not even mentioned. Rather, socio-historical factors affecting textual production, transmission, and reception generated this chequered group of texts.
Despite the obstacles, the compilers devised a detailed sixfold division of the Miscellaneous Schools and Writers whereby texts were organized not on the basis of their subject matter or stylistic conventions but rather of their ascribed function-for what end and how texts were to be used. Bibliographic subgroups were premised on whether texts were employed to establish sayings, authenticate claims, provide commentaries bringing together exegetical accounts, investigate the principles of objects, topically organize texts of the past, or collate together various writings.76 What the Miscellaneous Schools and Writers lacked of its own, the editors managed to impose by creating a structure where the ascribed function of texts guided the way in which they were grouped together and held apart.
The problem was also present in the Miscellaneous Records (zashi 雜史). Again, the compilers remark that the texts are disparate and numerous, making it difficult to order them.77 But the Miscellaneous Records have a common feature in that they either include the development of a specific event but do not provide information about an entire era, or merely narrate hearsay from one period and only from one writer's private accounts.78 The partial and often unreliable collection of sources was perceived as the defining criterion distinguishing the Miscellaneous Records from the Official Records and the Separate Records (zhengshi 正史 and bieshi 別史, respectively).79 The material factors of textual production are placed centre stage in the conceptualization of bibliographic order. Features above and beyond subject matter or stylistic conventions were at the heart of the formation of the Miscellaneous Records.
The conception of miscellaneous groups and the solutions given to the problems they posed reflects the general rationale of the order of knowledge: socio-historical features determining the production, transmission, reception, and ultimately use of texts determined the formation of bibliographic groups. The order of knowledge took into serious consideration extra-textual features in the formation of bibliographic categories.80 How a text was believed to have been composed, what kinds of sources it employed, how it was handed down from one generation to the next, and in what ways it could be put to function for the purposes of scholarship and governance were definitive in the formation of order.
These features are unrelated to the 'inherent' intra-textual properties of texts-such as content, style, genre, or disciplinary identity-which have by and large been the determinants of modern bibliographic orders and mistakenly extended to the Complete Writings. The way in which a text was classified depended on how it was envisioned to be employed. And how a text was employed was informed by the ways in which it was produced, transmitted, and received. The conceptualization of order did not depend on information found within texts but rather on information about them.
The construction of bibliographic categories followed the principle of socioscholarly function. For example, the editors note that the Edicts and Memorials (zhaoling fengyi 詔令奏議) section was in past compilations placed at the end of the Collections.81 They explain, however, that in the Complete Writings the section has been placed in the Records where they can provide clear testimony to the section. Furthermore, they note that the way in which the Edicts and Memorials is divided into its components is done with the intention of facilitating their reciprocal examination with other historical sources.82 The same rationale is seen in the Book of Rites (li 禮) section, which is divided into six groups chronologically arranged in order to allow for the comparison and examination of the source and flow of each group.83 And exactly the same justification is provided in the Lesser Sayings (xiaoshuo 小說) section, which divided texts into three streams.84 The reason these three groups are established, the editors note, is so that the diverging flow of the Lesser Sayings can be traced, showing how the execution of specific scholarly tasks was instrumental to the construction of bibliographic divisions.85 Bibliographic categories were 80 The practice of including 'extra-textual' features in the formation of bibliographic groups seems to predate the Complete Writings and, following Nicoll- established and positioned on the basis of their ascribed scholarly function, their capacity to assist in the execution of fundamental scholarly tasks.
The Masters preface opens with a general remark on the included texts' scholarly function that beyond the Six Classics, all texts which establish theories (lishuo 立說) belong to the Masters.86 Yet, the preface proceeds to order the sections of the Masters on the basis of their relevance to governance and the first six sections are grouped together on the basis of their significance in this regard. The Confucians (rujia 儒家) are placed first and are followed by the Militarists (bingjia 兵家) and the Legalists ( fajia 法家) whose knowledge was deemed essential to good governance.87 The Agriculturalists (nongjia 農家), Medicinal Schools (yijia 醫家), and Astronomical and Counting Methods (tianwen suanfa 天文算法) come next since these are vital to people's survival.88 These six sections are tied to the effective administration of the empire and their sequence mirrors their (ascribed) necessity in governance.
They are followed by two sections from which the lesser daos-skills or methods-can be observed, namely the Divinatory Arts (shushu 術數) and the Creative Arts (yishu 藝術).89 The next group, composed of four sections, constitute comprehensive resources for consultation. Their sequence-Treatises (pulu 譜錄), Miscellaneous Schools and Writers, Encyclopedias (leishu 類書), and finally the Lesser Sayings-was intended to capture their value. The final group are the external teachings (waixue 外學), placing the Buddhists (shijia 釋家) and Daoists (daojia 道家) texts last.90 The pattern reveals the collection's worldly orientation and its intended purpose to serve as a guide to governance. Identifying sections as tied to rulingand placing them first-reflects the compilers' understanding of texts relative to their social function in attaining sociopolitical goals. Artistic refinement, broad erudition, and 'external' (non-Confucian) cultivation were perceived as secondary to the state-oriented rationale of Confucian praxis.
The two-legged conception of order-social and scholarly function-was conceptualized as a unity. In the Masters preface, the editors remark that a scholar can uphold the rights and wrongs of the world by examining the principles in the Classics and can understand past and present successes and failures by studying the subtleties of the Records.91 The texts of the great Confucian scholars-and arguably, by extension but to a lesser degree, the texts in the Masters at large-illuminate dao and establish sayings which can function as comprehensive reference to the Classics and Records.92 This summary overview of the ascribed use of the texts shows how socio-scholarly function was thought of as one: a learned scholar became effective in his administrative duties through scholarship and a successful official expanded his knowledge through enacting the teachings of the sages. Texts were placed together and held apart based on how they were to be used by scholar-officials to attain intellectual as well as administrative objectives.
The significance of the socio-scholarly function as an organizing principle is highlighted by the prominence of hierarchical order. Time and again, the compilers separated texts which they understood to be on some level similar on the grounds that they were sufficiently distinct in their value. This is most acute in the Classics, where it is almost by definition true that texts which cohere with those of the other three sections are placed here solely on the basis of their significance. But the rationale extended to all kinds of writings: texts on different kinds of music, poetry, or even agricultural products, were ordered based on the ascribed value of their topics.93 It is not surprising that the Complete Writings had a hierarchical order. This is not because its makers were profoundly concerned with upholding 'orthodoxy' . Rather, it is because all knowledge organization structures have an in-built concern for value, whether acknowledged or not.94 The influence of hierarchical thinking has not been as overt in early modern and modern western projects because their concern with 'unearthing' shared properties between texts gave them a sense of objectivity, allegedly free of normative judgement. As Wilson-Lee has noted, however, the order in which humans took precedence over animals, men over women, and Europeans over others reinforced the belief that these hierarchies were natural and inevitable.95 There is no doubt that the editors of the Complete Writings carried their own biases. They certainly believed in the correctness of their division but not because they held it as an objective and factual description of the knowledge universe but rather because it provided an appropriate guide through it. Bibliographic groups were not a vehicle to capture internal properties that 92 Ibid. 93 For example, texts on bell-pipes from antiquity were placed in the Classics but other kinds of music were dispersed in the Masters and Collections; poetry from antiquity and in particular the Eulogies of Chu (chuci 楚辭) were given a special position in the Collections; and texts on tea-which was considered a luxury and not as essential as grain stapleswere placed further down in the Masters than writings on agriculture. 94 Appadurai 1988. 95 Wilson-Lee 2018 texts putatively possessed but rather structures which assisted in learning and carrying out the knowledge encoded in them. The relation between users and texts-society and knowledge-thus emerges as the cornerstone of the bibliographic order.
Despite their differences, the Aristotelian and Wittgensteinian models conceptualize categories in terms of properties, placing exclusive focus on what objects are. Organizing in terms of function, however, constitutes a different mode of organization. Failing to recognize this fact is what makes it hard to understand-let alone appreciate-the knowledge structure of the Complete Writings.
Think of a toolbox. Inside there are nails, bolts, screws, and tools of different kinds. The Aristotelian and Wittgensteinian would organize the toolbox by considering the properties of tools: the former would place identical tools together in ever smaller categories, while the later would put similar tools in nearby compartments. The Complete Writings way, however, would examine how different tools are used-for what kinds of purposes-and place tools used together in the same section. It would examine how frequently tools are used and how important they are for different kinds of jobs. If some tools are used more often than others they will be placed not only together but at a point where they can be more easily accessed. Similarly, if certain tools are infrequently used or have a very limited function they will be placed deeper inside the box.
A function-oriented toolbox may appear messy to the untrained eye. To the person using the toolbox, however, there is a well-defined rationale. In the same way, the Complete Writings which might prima facie appear to be lacking a clear order is in fact sensible for its makers and intended users. The reason why the Complete Writings may appear haphazard today has to do with the fact that the kinds of questions professionalized, modern university academics try to answer are different from those of eighteenth-century scholars in China. The properties-based and function-oriented ways of organizing toolboxes, libraries, or the world create vastly different arrangements. Taking the Complete Writings seriously as a scholarly project offers a comprehensive and systematic conception of order which challenges-and therefore has the potential to enrich-modern conceptions of collecting and organizing knowledge.

Conclusion
The Complete Writings would not be the collection it is were it not for its organization. Without a clear classification, it would only be an accumulation of writings and would not achieve any of its purposes, from the simple task of allowing users to locate texts to the more complex goal of acting as an advising guide. Given its capacity to define what is important and unimportant, the collection was no doubt an exhibition of imperial power.96 However, as I have suggested throughout this article, the collection was also-and perhaps just as much, if not more-an exploration of the nature of knowledge and its relationship to systematic classification. Boosting scholarship, refining its methods, and improving its categorical structure were unmistakably essential to the project. The question of order within the Complete Writings opens up a new way of exploring the nature of collecting in late imperial China. The late Ming and high Qing saw an outburst in the publication of collections: content varied from dreams to sutras and publication was both imperially sponsored and privately financed for an expanding market.97 While contemporary scholarship has taken note of this veritable explosion, the question of how these collections were organized has remained marginal. Yet, as this article has shown, exploring the classification of any one collection offers insights into a far broader set of issues, from perceptions of the natural world to the configuration of social structures. Exploring the order of knowledge can bring us closer to the lived reality of scholars in the late imperial era since, as the celebrated scholar Wang Mingsheng 王鳴盛 held, "in learning, bibliographic studies are of outmost importance and it is necessary to apply oneself here first in order to be able to find one's way" (目錄之學學中第一要事必從此問塗方能得其門而入).98 The history of classification practices can illuminate the more generalized passion for collecting, the interplay between imperial and private collections, and the role collections had in consolidating or challenging received frames of knowledge.
Chinese classification practices constitute an untapped pool of experience for contemporary collections, exposing deep-rooted assumptions and suggesting alternative ways of collecting and organizing. Our conceptions of post-disciplinary arrangements of knowledge, the effects of digital media on classification, and the rise of information and data management only stand to benefit from exploring and integrating the organizing methods, aims, and conclusions of late imperial collections. No matter where or when, collecting has been tantamount to ordering. This simple fact, which reflects our shared 96 Guy 1987: 78. 97 Feuerwerker 1976Elliot 2009: 118, 164. Also, see Inaba Ichiro 稲葉一郎 2006 for the importance of imperially sponsored projects. See Chen 2008 for his encyclopedia of dreams. 98 Quoted in Wilkinson 2015: 936.