The Case of “Xi shuai” 蟋蟀 (“Cricket”)

The widely discussed corpus of looted bamboo manuscripts now in the possession of Tsinghua University (Beijing) contains a short text of fourteen bamboo slips—five of them broken and incomplete—that on the back of its final slip is named Qi ye 耆夜,Footnote ¹ a title that appears to refer to a banquet in celebration of a military victory over the state of Qi in the initial years of the Western Zhou dynasty (1046–771 b.c.e.). The manuscript, written in the Warring States period script of the southern state of Chu and believed to come from a southern tomb, has been tentatively dated to around 300 b.c.e.Footnote ² The brief manuscript text narrates the process of a drinking ritual, in which first King Wu 武王 and then his brother the Duke of Zhou 周公 each present two toasts, each toast accompanied by the performance of a short song. The narrative ends with the text of the Duke's second poem, titled “Xi shuai” 蟋蟀 (“Cricket”), a poem that the Duke either extemporized or sang from memory.Footnote ³

This poem in three stanzas is very closely related to a poem of the same title (Mao shi 114) in the “Tang feng” 唐風 section of the Mao shi 毛詩, the received version of the ancient anthology of the Classic of Poetry. According to Han sources, the anthology was compiled by Confucius when he reduced a corpus of “more than 3,000 pieces” (sanqian yu pian 三千餘篇) by “removing duplicates” (qu qi chong 去其重) to just 305, all of which he sang to string tunes (jie xian ge zhi 皆弦歌之).Footnote ⁴ If the manuscript is authentic, then it represents one of only two cases in early Chinese literature (the other one also in the Tsinghua manuscripts)Footnote ⁵ where a multi-stanza poem (or some version thereof) from the Poetry appears in full outside the anthology itself. While the various received texts from early Chinese literature—both historiographic and philosophical—contain numerous quotations from the Poetry, there is only instance, in Guoyu 國語, where a full poem is quoted, “Haotian you cheng ming” 昊天有成命 (“Grand Heaven Had Its Accomplished Mandate”; Mao shi 271), a song of merely 30 characters.Footnote ⁶ The longest quotation from the Poetry in any received early text is a single stanza of 48 characters from “Huang yi” 皇矣 (“Great Indeed”; Mao shi 241), found in Zuo zhuan 左傳.Footnote ⁷ By contrast, the manuscript text appears to have contained three stanzas of “Cricket” of altogether 132 characters, of which 108 are still fully legible while twenty-four others are lost to damage.

This article presents a brief comparison of the two versions of “Cricket” and then proceeds to discuss a number of basic methodological issues concerning the study of the manuscript version. The main part of the article is concerned with the nature, composition, and transmission of poetry before the dawn of the Chinese empire.Footnote ⁸

About half of the text is different across the two versions. Compared to the received text, the manuscript “Cricket” shows the following differences:

1. 1. Its many textual variants include orthographic variants where different characters are used to write the same words (sounds); lexical variants where different but semantically similar words are written; and lexical variants that differ in meaning, in some cases constituting entirely different lines.

2. 2. Its perspective of speech is different in that it does not include the first-person pronoun wo 我. Instead, descriptive sentences attributed to “us” in the received text are addressed as exhortations to “[you] noble men” in the manuscript.

3. 3. Phrases are transposed from one line to a different one.

4. 4. It contains two additional lines per stanza, but not by the simple addition of a couplet. Instead, parts of these additions also exist in the received version, but in contracted form.

5. 5. The final couplet of each stanza has four additional characters, thus violating the tetrasyllabic meter. However, the meter of these couplets can still be reconstructed as tetrasyllabic by removing unstressed particles.Footnote ¹¹

6. 6. It rhymes differently.

7. 7. It is placed in an entirely different historical context.

To begin with the final observation, here is the Mao shi Preface to “Cricket”:

Here, the Mao shi preface not only dates the poem three centuries later than the manuscript does, but also foregoes any specific notion of authorship. Moreover, it regards “Cricket” as a poem of criticism, while the manuscript views it as a celebratory song. Together, the preface and the poem—the latter through the differences in form and content listed above—raise fundamental issues for our understanding of the composition, transmission, meaning, audience, and purpose of poems such as “Cricket.” In the following, I explore some of these issues. First, however, it is important to address the problem of looted manuscripts and how it severely limits our analysis.

The Problem of Looted Manuscripts

To date, the Shanghai Museum (manuscripts acquired in 1994), Yuelu Academy (2007), Tsinghua University (2008), Peking University (2009, 2010), Anhui University (2015), and possibly others are in the possession of large caches of unprovenanced—recently looted and purchased—bamboo manuscripts from, presumably, Warring States through Han times. None of these manuscripts has been sufficiently and independently tested, but all are claimed to be authentic by the institutions that own them and the scholars who publish them. Within this situation we have not one problem but three: the problem of possible forgeries, the problem of lack of contextual information, and the problem of how to work with stolen goods in ethical and responsible ways. I will briefly address them separately.

According to the Tsinghua editors, the manuscripts were donated by a Tsinghua University alumnus in 2008, after they had surfaced on the Hong Kong antiquities market. The editors give a date of 305 b.c.e. +/– 30 years, based on carbon-14 dating of a single bamboo slip.Footnote ¹² While most researchers accept the authenticity of the Tsinghua manuscripts, at least one group of scholars claim that the Qi ye manuscript is a modern forgery.Footnote ¹³ Whether or not one agrees with their (so far isolated) conclusion, it must be noted that the ink of the actual characters has never been tested, and that blank, unwritten bamboo slips from ancient sites exist in considerable quantities. It does seem possible that a modern person—or a collaborative team—deeply familiar with both ancient Chinese texts and Warring States paleography and phonology could produce a Chu-script text like those found among the Tsinghua manuscripts. Claims that such forgeries would be too difficult or impossible remain impressionistic and cannot be proven or disproven. A poor forgery—the widely shunned manuscripts owned and published by Zhejiang University may be a good candidate—does not disprove the abilities of potential forgers; if anything, it only reveals a set of criteria easily known and observed by them. We can never know the skills of the most capable forgers because they remain undetected, but we underestimate them at our own peril. As the most prestigious academic and cultural institutions in Mainland China are now in an arms race to purchase ever more important, more beautiful, and more complete manuscripts on the Hong Kong antiquities market, prices will rise in response to demand, and with them the incentives to create highly sophisticated forgeries. As we study the looted manuscripts available so far and report with great excitement our new insights derived from them, we believe them to be authentic, and we surely hope so and want them to be so, but we do not know.

Even if the manuscripts are genuine—looted but not forged—they present enormous ethical, legal, and scholarly problems that have barely been discussed.Footnote ¹⁴ At this point, the market increasingly encourages and rewards more looting and illegal selling. In recent years, large collections of manuscripts have been archaeologically recovered, but—unlike with the earlier discoveries of the Han dynasty tombs at Yinqueshan 銀雀山 (Linyi, Shandong, excavated 1972), Bajiaolang 八角廊 (Dingxian, Hebei, 1973), Mawangdui 馬王堆 (Changsha, Hunan, 1973), or Shuanggudui 雙古堆 (Fuyang, Anhui, 1977)—most of their contents are now administrative, legal, or technical in nature, as opposed to texts related to the received classical tradition of literary, philosophical, or historical writings.Footnote ¹⁵ By contrast, most looted manuscripts that have been acquired over the past twenty-five years—the major exceptions so far being the Qin dynasty administrative, technical, and legal texts held by the Yuelu Academy and the Qin technical manuscripts held by Peking UniversityFootnote ¹⁶—are precisely that: philosophical, literary, or historical texts with received counterparts, or texts of this kind that are more or less directly related to the classical tradition. The situation is particularly grave with regard to pre-imperial (as opposed to Qin and Han) manuscripts: no recent archaeological find has yielded even a fraction of the Shanghai Museum or Tsinghua University texts; even the last major such find, at Guodian 郭店 (Jingmen, Hubei, 1993), is dwarfed by them.

As a result, in working with philosophical, literary, or historical manuscripts we now depend overwhelmingly on looted artifacts as our benchmarks in studying and rethinking early Chinese intellectual and cultural history. Neither the Shanghai Museum nor the Tsinghua University texts provide any kind of context: we do not know from where these manuscripts come, whether or not they are complete, to whom they may have belonged either personally or institutionally, who produced them, who their audience was, what purpose they may have served, what their place was in relation to a possibly much larger body of other (tomb?) artifacts, how they were originally arranged in their physical surroundings, and so on. Instead, we are facing displaced and nearly disembodied texts and ideas whose only remaining connection to their original social context is their own material and visual features. Even if all our looted manuscripts turn out, as we hope, to be genuine, they do not lend themselves to conclusions about their actual Sitz im Leben. In consequence, studies on these manuscripts focus mostly on their paleography or on the ideas expressed in them. Yet lacking all context in terms of origin, circulation, and reception, it is impossible for us to reconstruct their place in intellectual history: any such manuscript could be a possible textual witness within a series of other witnesses, or it could just as well be a random, isolated, and idiosyncratic piece of writing signifying little beyond its own existence. We simply do not know, and we must acknowledge this lack of understanding before advancing generalizations regarding the intellectual history of early China. We also cannot draw conclusions from isolated pieces of evidence across different types of writing: particular practices of text copying of legal, technical, or administrative texts, for example, may have depended on specific institutional structures and demands that, on the other hand, may not have applied to literary or philosophical texts at all. In fact, generalizations may not even be drawn across different texts within a single corpus of looted manuscripts because we do not know whether all these texts indeed originated from a single environment or were cobbled together before they were sold in Hong Kong. Strictly speaking, each looted manuscript can only be studied within its own boundaries; and often we do not even know what these boundaries may have been originally.

Beyond these basic scholarly issues with looted manuscripts are, finally, also ethical and legal ones. They implicate every scholar, Chinese or foreign, who validates them through academic research: more recognition generates more demand, and more demand stimulates more supply, that is, more looting. Scholars of early China are well aware of this dilemma: on the one hand, principled arguments exist against participating in this vicious cycle. On the other hand, the historical knowledge possibly gained from some of these manuscripts—even within the limits outlined above—may be too important to ignore. In their accumulation, these newly available sources have the potential to change and expand entire narratives of early Chinese history; to ignore them would mean to willfully perpetuate traditional views that we know are in urgent need of modification. Does the quest for historical knowledge constitute a defensible ethical stance in accepting looted artifacts? Even if so, could it also be ultimately self-defeating as it may encourage further looting, with the profound loss of knowledge, outlined above, as the inevitable result?

In my view, there is no single clear answer to these questions. At this point in the history of our field, each scholar must conscientiously choose and defend their position, and these individual choices must be accepted. What I would like to suggest, however, is the following: the destruction of knowledge that occurs when manuscripts are looted from ancient sites deprives these sites, together with their objects in situ, from taking the place they deserve in world cultural heritage. Looting is theft from humanity in general and from Chinese civilization in particular. For us, scholars and curators of the ancient past, Chinese and foreign, perhaps the time has arrived to come together, acknowledge the dilemma we share, speak out collectively, and urge the authorities at every level to double and triple their efforts in fighting the looting. The emergence and subsequent purchase of yet another corpus of looted manuscripts—even under the name of repatriation—is not cause for celebration. It is cause for our collective and public concern in defense of Chinese cultural heritage.

One Poem or Two? Questions About the Composition and Transmission of “Xi shuai”

The majority of studies of the Qi ye manuscript “Xi shuai” poem that I have consulted all ask the same question: are the manuscript “Xi shuai” and its received Mao shi counterpart two separate poems or just one? Nearly all studies contend it is just one poem, albeit in two versions (without defining the line that separates a “version” from a different text). This inevitably leads to a second question: which of the two is earlier? Implied in this question is further: which one is the “original,” or at least closer to some even earlier “original version,” and therefore also more trustworthy as a representation of how this particular poem looked in the beginning? And how should one date the earlier poem? The scholars cited aboveFootnote ¹⁷ have come to different conclusions: most of them believe the manuscript version is the older one; only Cao Jianguo 曹建國 sees the manuscript text as the work of someone who imitated the Mao shi poem. Li Xueqin 李學勤 has taken the most extreme position in claiming that the manuscript version reflects the original, early Western Zhou composition by the Duke of Zhou. Other scholars privileging the manuscript text have been more cautious, claiming instead that it represents a Springs and Autumns or Warring States composition. Finally, Li Rui 李銳 has thoughtfully surmised that the two poems may have originated from a single text but then got separated into two mutually independent “textual lineages” (zuben 族本), and that it is therefore fruitless to ask which of the two texts is the earlier one.

The Mao shi “Xi shuai” appears as the more standardized text in several ways, most importantly in the final lines of each stanza. On the assumption that canonization also involves standardization, the Mao shi version may thus appear to be the later text, where—as noted above—the extended final lines of the manuscript text have been transformed without semantic loss into the canonical tetrasyllabic form.Footnote ¹⁸ But what would this mean for the canonization of the Poetry altogether? Did the Tsinghua manuscript stand somehow outside the canonization process, containing some idiosyncratic version of “Xi shuai” that by 300 b.c.e. was otherwise already replaced by the canonical form similar to the one found in the Mao shi? Or had, more generally, the redaction of the Poetry in its later-known form not yet occurred by 300 b.c.e., despite its traditional attribution to Confucius and regardless of the fact that in two bamboo manuscripts from Guodian, also dated to c. 300 b.c.e., the Poetry is listed as part of the classical curriculum of the “Six Arts” (liu yi 六藝)?Footnote ¹⁹ In the latter case, the Mao shi version must be very late indeed, and the result of a graphic and phonological standardization for which William H. Baxter has called it “a Zhōu text in Hàn clothing: both its script and, to some extent, its text have been influenced by post-Shījīng phonology.”Footnote ²⁰ Yet under either scenario, we cannot take the received Mao shi as altogether representative of the Poetry at the time of the Tsinghua manuscript.

What many current readers of “Xi shuai” share, though they do not always make this explicit, is the presumption of a single original moment, and single author, of textual composition that resulted in a written Urtext from where the poem began its journey though time along relatively stable lines of manuscript copying. In the Western tradition, this is the model familiar since at least the fourth century b.c.e. when the Alexandrian philologists began to create editions of the Homeric epics where they had to decide between different textual variants found in different manuscripts of the same text.Footnote ²¹ It was the great German philologists Friedrich August Wolf (1759–1824) and Karl Lachmann (1793–1851), however, who pioneered the scientific method for establishing critical editions of classical texts.Footnote ²²

In one way or another, scholars operating on the assumption of direct copying from manuscript to manuscript as the principal form of textual transmission subscribe—whether they know it or not—to the Lachmannian model of the stemma codicum, where written texts begin from a single point of origin and from there branch out into separate lineages, an idea that, for example, is made explicit in Li Rui's notion of different zuben. In 2002, I offered the following alternative reflection:

Today, I find it no longer meaningful to suggest that “all these versions go back to an Urtext”; instead, I would now emphasize even more strongly that “there is no single written original behind the different versions” of a poem. We do not know how to imagine the moment when someone composed the first “Xi shuai” poem, or how any other such poem subsequently derived from it. In thinking about the two “Xi shuai” poems, the question is therefore not which of them might be earlier and which later, nor can we embark on a search for “the original poem.” Instead, we may address the relationship between the two by asking: under what model of textual transmission can we explain how the poem developed from the Mao shi version to the manuscript version, or vice versa? Or should we rather reject any such linear development from one to the other?

A simple model of consistent visual copying from a written original cannot explain why one text has more phrases than the other, why shared phrases and entire lines appear in different places in the two poems, why the two poems differ in their perspectives of speech, why they have different rhymes, why they are contextualized and historicized in different ways, and why they contain phrases and lines of entirely different content. Whatever textual practices led to the two texts we now have (and possibly numerous others we no longer have, or have yet to discover), they must have included acts beyond those performed by faithful scribes. While individual scribes in both medieval Europe and Tang dynasty China, “in the absence of an ideological conception of an original that has a unique, authorizing form, felt free to produce copies that varied from their exemplars at virtually every level,”Footnote ²⁴ the differences between the two “Xi shuai” texts reflect something more profound altogether: two parallel, separate realizations of a text that are not simply written differently, but—far more consequentially—are writing something different altogether, namely, texts with different historical contexts and meaning. Here, the question of textual reproduction becomes one of textual transformation and renewed composition that involved something close to authorial agency even though it was performed on some preexisting model. Nothing of this is primarily a question about written or oral transmission; it is a question about conceptions of textual fidelity, fixidity, integrity, authority, and control. What can be ruled out for the two “Xi shuai” poems, however, is clear: there is no direct line from one version to the other where a scribe copied the former from the latter, if by “copied” we mean an attempt to reproduce an existing model with some degree of fidelity. At least one of the two people who wrote the two poems knew of another version—or someone before him had known of it—but chose to compose something new and different. Who these people were, what their overall education was, to which social stratum they belonged, and to what extent they participated not merely in scribal practices but also in the shaping of poems like “Xi shuai” and other literary texts, we have no idea.

However, from the exquisite calligraphy and generous spacing of some of the literary, philosophical, and historical Warring States manuscripts, we may conclude that they were as much meant to be seen as to be read. Whoever wrote the Qi ye manuscript was trained not just in the writing system but in the execution of beautiful, representative calligraphy. The manuscript was probably valued for its appearance and was possibly created as a display item that conferred prestige on its owner. Equally probable is that such an artifact of writing was not a primary text but a secondary one: among the cultural elite, at least the local elite, its contents will have been known and cherished already.Footnote ²⁵ In general, this will have been true for manuscripts placed in tombs, and even more so for manuscripts made specifically as burial items. The latter must have existed in an earlier form before, and they were familiar, at a minimum, to the person buried with them. At the same time, we do not know how much authorial or editorial agency we can grant the scribe of such a manuscript. What William G. Boltz has noted for the early Laozi 老子 materials is just as true for poetry:

Altogether, at stake with the two “Xi shuai” texts are fundamental questions like these: what actually is the textual identity of a poem in early China? How does it come into being? How shall we think about its authorship? Who owns or controls a poem? What are our most plausible models for textual transmission and circulation?

Repertoires, Composite Texts, Authorship

In an influential 2005 article, Boltz advanced the notion of “composite texts” made of “building blocks,” that is, smaller textual units that could be flexibly compiled in different contexts.Footnote ²⁷ This notion fits the available evidence and renders concepts such as “the original text,” “the original author,” or “the time of the poem's composition” meaningless. Unlike Greek or Roman literature—but very much like most other ancient and medieval literary traditions—early Chinese literature before the advent of the imperial state did not emphasize the notion of individual authorship, ownership, or control of texts. Within the Poetry, even in poems such as “Grand and Lofty” (“Song gao” 崧高; Mao shi 259) and “The Multitudes” (“Zheng min” 烝民; Mao shi 260), which in their final quatrains declaim “Jifu made a recitation, its lyrics are very grand” (Jifu zuo song, qi shi kongshuo 吉甫作誦，其詩孔碩) and “Jifu made a recitation, may it be gentle as the clear wind” (Jifu zuo song, mu ru qing feng 吉甫作誦，穆如清風), such utterances are not statements of authorship but references to performances that seem external to the poems themselves.Footnote ²⁸ Likewise, a text such as Confucius’ Discussion of the Poetry (Kongzi shilun 孔子詩論), a looted manuscript text so titled by its modern Shanghai Museum editors,Footnote ²⁹ discusses a great many titles from the Poetry and yet never once refers to a poem's moment or purpose of original composition, or to the identity of its author. Probably a teaching manual for how to use poems and their topics in situational contexts, it focuses instead on their core ideas, often captured in a short descriptive phrase or even single defining word.Footnote ³⁰ Finally, and more generally, not one of the literary, philosophical, or historical manuscript texts that have surfaced—looted or excavated—over the past four decades carries the name of its author, nor does any such manuscript name a specific author of any other text.

As in the medieval poetic traditions of China and Europe,Footnote ³¹ the general lack of emphasis on authorship in early Chinese poetry is directly related to the overall instability—mouvance in Paul Zumthor's terminology, variance in Bernard Cerquiglini's—of poems. In all these traditions, the “author function” (Michel Foucault)Footnote ³² does not exist as a controlling agency in either the interpretation or the stability of the text, and any effort to retrospectively “reconstruct” or “discover” a particular author or specific historical moment of composition is not merely doomed to fail but is fundamentally misguided in the first place. This is because poems existed not as discrete individual entities but as ever-renewed instantiations of “poetic material” or “repertoires” (Stephen Owen's terms) of both themes and expressions. Consider Owen's description of early medieval Chinese poetry:

This is precisely what we see in numerous examples of early Chinese texts, including—but not limited to—poetry. Such fluidity, which Gregory Nagy has also claimed for the Homeric epics,Footnote ³⁴ is related to a notion of poetic composition and recomposition in ever renewed contexts and on ever new occasions, be they in oral performance (Zumthor) or in the creation of new manuscripts (Cerquiglini). In other words, the stability of texts is not controlled by acts of visual copying, nor are texts, as in the Lachmannian model, to be lined up on a chronological axis. Instead, there is a constant multiplicity of different textual realizations derived from a circumscribed body of poetic material and expressions (see below). The elegance of this understanding of ancient and early medieval textuality is easy to grasp: first, it explains better than any other model the numerous textual examples both within a number of different cultures and between them; second, it transcends the simplistic dichotomy of “oral versus written” and eschews unfounded ideas about the stability of written texts, or even the desire for such stability;Footnote ³⁵ and third, it places meaningful limits on the use of general notions such as “building blocks” and “composite texts.”

Let us now return to “Xi shuai” and consider once more the differences between the two versions: one text has more phrases than the other; shared phrases and lines appear in different places in the two poems; the two poems have different perspectives of speech; they have different rhymes; they are contextualized and historicized in different ways; and they contain phrases and lines of entirely different content. I believe that none of this is most plausibly explained by a process of direct visual copying, nor does it tally with presumptions of individual authorship, or of textual stability of written artifacts over time. Instead, the two “Xi shuai” texts are related through their overall theme, their images, and circumscribed sets of expressions that clearly distinguish them from any other poems: they are mutually independent instantiations of a shared repertoire, or “poetic material.” It is easy to imagine how other “Xi shuai” poems, were they to be discovered, would likewise be different. What does not change in either “Xi shuai” text is their fundamental idea, that is, the appraisal of pleasure in moderation.

Yet even this fundamental idea was apparently subject to variation. The Shanghai Museum manuscript Confucius’ Discussion of the Poetry contains repeated series of definitions that appear to capture for each poem its core idea, or rather the idea for whose expression the poem in question may be the one most appropriately applied. The passage involving “Xi shuai” reads as follows:

Assuming that both the Tsinghua and the Shanghai Museum manuscripts are genuine and come from roughly the same time and geographical region—their regional identity being indicated by their shared Chu-area script—how does this comment on “Xi shuai” square with either the received text or the one in the Tsinghua Qi ye manuscript? It does not.Footnote ³⁷ Instead, we are facing yet another, completely different reading of “Xi shuai”—one that would be incompatible with the historical contextualization of either the Tsinghua or the Mao shi text. “Understanding difficulty” may therefore refer to yet another poem under the same title, one that evoked the imagery of “Xi shuai” while conveying a message of both “difficulty” and the awareness of it. This poem we do not know; what we do recognize, however, is the overall lack of a unified “Xi shuai” poem. In other words, there is no single poem; there are various realizations of poetic material under the title “Cricket” that took shape as multiple parallel, mutually independent texts, allowing for multiple hermeneutic possibilities. The title “Xi shuai” does not signify a single poem; it signifies a multiplicity of poetic expressions or, more precisely, a poetic discourse from which multiple different expressions could be generated. These expressions could exist parallel to one another, defined not by the existence of some reified “original poem” but, instead, driven by a range of situational contexts, demands, and expectations. The individual poem, if such a concept is even appropriate, was not generated by some original authorial intent but, to the contrary, arose in ever new form under specific circumstances.

If this is true of “Xi shuai,” it will also be true of other texts in the Poetry. The principles of textual reproduction transcend any particular example as they reflect the larger socio-cultural environment where early Chinese poetry unfolded and took its place. It should therefore be possible to identify other repertoires of poetic material and how these repertoires were realized in particular poems. Given that the model of textual production invoked here is predicated on flexible situational adaptation—as opposed to mechanical reproduction—we should be prepared for a wider range of possible realizations of poetic repertoires.

Indeed, examples are easy to find. The first is the short poem “Jing zhi” 敬之 (“Be Reverent”; Mao shi 288) that in the Mao shi comprises 54 characters.Footnote ³⁸ The Tsinghua manuscript version of 55 characters, found in the text Zhou gong zhi qinwu 周公之琴舞 (The Zither Dance of the Duke of Zhou), not only shows many of the same kinds of differences that are visible between the two “Xi shuai” poems, including entirely different phrases and lines, lines of different length, and an additional line. But two features are particularly interesting: first, the manuscript text is explicitly divided into two halves, with the first being called an “opening” (qi 啓) and the second a “coda” (luan 亂), perhaps suggesting a shift in musical performance (note that the poem is introduced as a “zither dance”). The second feature that distinguishes the poem in the manuscript text is the fact that, unlike its counterpart in the Mao shi, it does not stand alone: it is only the initial section, or musical movement, of a nine-part “zither dance” suite that is explicitly labeled as such. If this is one form in which “Jing zhi” was known as late as around 300 b.c.e.—not as an individual poem but as part of a larger suite with a different title—then its binding canonization as part of the “Eulogies of Zhou” (“Zhou song” 周頌), presumably the earliest section of the entire Mao shi anthology, had yet to be accomplished.

Before discussing further examples of poetic repertoires as the source of individual poetic utterances—including what then became discrete poems in their own right—it is necessary to reflect on the possible boundaries of such repertoires as well as on the range of possible variation. In an essay published in 2005,Footnote ³⁹ I calculated the percentage of variants between the Mao shi on the one hand and six pre-imperial and early imperial manuscripts with text fragments from the Poetry on the other. Of the six manuscripts, two were from Guodian and two others were part of the Shanghai Museum corpus of looted texts (the Tsinghua texts were only published years later). The total number of characters of Poetry quotations in these four (presumably) pre-imperial manuscripts is 464, and the total number of variants (of any kind) is 181, or 39.0 percent. It should be noted, however, that of the 464 characters, there is significant overlap between the Guodian Black Robes (Ziyi 緇衣, with 193 characters from the Poetry) and its near-identical counterpart in the Shanghai Museum corpus (with 157 such characters). In other words, of the 464 characters from the Poetry in pre-imperial manuscripts before the discovery of the Tsinghua manuscripts, 350 come from what is essentially a single text: Black Robes, with more than half of these characters in duplicate. Moreover, this text as a whole, including the Poetry fragments, has a received counterpart in the Liji chapter of the same title.Footnote ⁴⁰ Thus, Poetry quotations in previously unknown texts—the Five Modes of Conduct (Wu xing 五行) from Guodian and the Confucius’ Discussion of the Poetry from the Shanghai Museum corpus—amount to merely 114 characters, of which 44 (38.6 percent) are variants, if compared to the Mao shi.

In a further step, I then calculated the percentage of variants between the Mao shi and the reconstructed fragments of the Western Han “three lineages” (san jia 三家) versions of the Poetry for the same set of Poetry fragments found in manuscripts. Here, instead of the average variant ratio of about 39 percent between the Mao shi and the manuscripts, the ratio is only 5.23 percent. What do these data suggest? To begin with, we may be overstating the number of manuscript variants by misconstruing the significance of particular graphs. Specifically, Chu manuscripts such as those from Guodian or in the Shanghai Museum corpus represent a distinct regional writing style whose characters do not always correspond to the standardized kaishu 楷書 forms known from the received tradition. In other words, a Chu character may appear to be a variant when compared to its Mao shi counterpart but in fact may simply represent a particular regional way of writing. Such slippage alone, however, cannot account for a variant ratio in the manuscripts that is seven to eight times higher than what the transmitted san jia variants show, nor does it explain the variants between different preimperial manuscripts in Chu script. Some other forces must be in play.

The so-called san jia fragments, attributed to the exegetical lineages of the Qi, Lu, and Han versions of the Poetry anthology, were culled from nearly thirty transmitted texts of early China, but they are not the only Poetry fragments altogether; they are a distinct minority. A far larger group of Poetry parallels and quotations—those consistent with the Mao shi—can be found across many more early texts preserved in the received tradition. Yet considering the very high ratio of variants in manuscripts—including between manuscripts from the same geographical region and historical period—it is inconceivable that Poetry citations across a wide range of pre-imperial texts from different centuries and places would have accorded consistently with the received Mao shi. Instead, they must be the result of retrospective (Han and post-Han) standardization.Footnote ⁴¹

But what about Poetry fragments and quotations in recently discovered manuscripts? If the received Mao shi poems were but one of many actualizations of larger poetic repertoires, should we not expect to find a larger range of lines that do not match the particular Mao shi poems? According to my earlier analysis, the high variant ratio in manuscripts, compared to the Mao shi, overwhelmingly reflects differences in writing, but not differences in sound: while words were written with different graphs, phonologically these graphs are close enough to one another to be interchangeable and hence simply reflect different ways of writing the same (or nearly the same) sound, which in most cases, presumably, would also have been the same word.Footnote ⁴² In short, the fragments and quotations from the Poetry by and large cohere phonetically—and hence probably also semantically—with the received text of the Mao shi. Would this general coherence not suggest that the individual poems of the Mao shi were indeed the same as those that are quoted in manuscripts?

This question is useful for delineating the range of variance and mouvance across the different actualizations of poetic repertoires. In my view, the available evidence puts clear limits on variance and mouvance—limits that are reflected well in the “Xi shuai” and “Jing zhi” poems—but it does not contradict the notion of repertoires. First, as can be seen in the “Xi shuai” and “Jing zhi” poems, even different realizations from the poetic repertoire still share a very significant number of phrases. Second, when a poem is quoted, it is rarely identified by title but, instead, is called upon with a more general signifier such as “the Poetry [or: a poem] says” (shi yue/yun 詩曰/云) or “the Major Court Hymns [or: a Major Court Hymn] says” (daya yue/yun 大雅曰/云). In all these instances, we do not know prima facie to which poem the quoted lines belong. The very fact that, in most cases, poems are invoked not by titles but by generic signifiers (the Poetry/the “Major” or “Minor Court Hymns,” etc.) itself suggests a looser concept of poetry that is not defined by individual titles of discrete poems. (Note here that there is no difference between the expression “a poem” and “the [anthology of the] Poetry.”) And third, references to and quotations from the Poetry are highly uneven in their distribution.Footnote ⁴³ Poems are not quoted randomly: some signature lines are quoted frequently, while most other lines are never quoted at all.

For example, the poem “Shi jiu” 鳲鳩Footnote ⁴⁴ (“Cuckoo”; Mao shi 152) comprises ninety-six characters in twenty-four lines.Footnote ⁴⁵ These are arranged into four stanzas of six tetrasyllabic lines each. Quotations of this poem can be found in about twenty-three different passages in received Warring States and Han texts,Footnote ⁴⁶ and in addition also in the Guodian and Shanghai Museum manuscripts. The three longest quotations—one in Xunzi, the other two in the Han texts Lienü zhuan 列女傳 and Da Dai Liji 大戴禮記—contain six lines of altogether twenty-four characters; all three quote exactly the same six lines, which means that these quotations themselves may be interrelated. In their vast majority, quotations from “Shi jiu” amount to one or two lines, and very rarely four. Altogether, of the twenty-four lines, only fourteen are quoted, that is, a total of fifty-six out of ninety-six characters. Of these fourteen lines, several are repeated within the Mao shi poem. Meanwhile, how is the poem quoted in manuscripts? Two lines are quoted in the Black Robes texts from Guodian and in the Shanghai Museum corpus; the same two lines are quoted in the Guodian and Mawangdui Five Modes of Conduct manuscripts (the Mawangdui text dating from before 168 b.c.e.); two more lines are quoted in the Shanghai Museum Confucius’ Discussion of the Poetry; and one more line is quoted in the Mawangdui Five Modes of Conduct. Altogether, only five lines from “Shi jiu” are quoted in these five different manuscripts.Footnote ⁴⁷ All five lines are part of the first stanza of the Mao shi poem and are also contained in the three longest quotations (of six lines each) of Xunzi, Lienü zhuan, and Da Dai Liji.

What does this mean? First, we cannot reconstruct any such poem from its quotations in either early transmitted sources or recently discovered manuscripts. And second, while a considerable number of quotations of lines appear in the received Mao shi, these quotations also are highly selective: across a wide range of texts, they tend to invoke the same lines over and over again—the signature lines with which a given poem is identified. These, of course, are not just the signature lines of that poem; they are the lines that express the core idea of the respective poetic repertoire to which the poem belongs. From the very concept of poetic repertoires, we would expect precisely what we see in these quotation patterns: a focus on a limited set of phrases that define the repertoire and are frequently repeated in most or all of its actualizations.

Let us look at “Shi jiu” more closely to consider why all manuscript quotations come from what in the Mao shi version is its first stanza. The poem, like so many of the “Airs of the States” (“Guofeng” 國風), is highly repetitive across its stanzas:

Lines 1 and 3 are repeated in each stanza. Line 1 triggers line 2: in the first stanza, line 2 numbers the young ones; in the following three stanzas, they are sitting in different trees. Line 3 is a statement on the virtuous person, which triggers line 4 in different ways: his “deportment” is “unfailing” or “flawless,” “his girdle is of silk,” “he rectifies the men of the state.” Line 5 of each stanza is anadiplosis, repeating line four; thereafter, line 6 concludes the stanza: “his heart is like tied,” “his cap is black-mottled grey,” “he rectifies the states of the four quarters,” “how shall he not have ten thousand years.”

In other words, the composition is modular; each stanza repeats the overall structure while introducing minor new elements in lines 2, 4 (repeated in 5), and 6. The single image taken from nature is that of the cuckoo and its young ones; the single person lauded is “the virtuous man, the noble man” who does what a virtuous, noble person does: he wears his ritual cap, is correct in his behavior, and rectifies others. To bring the poem to an end, the final line presents the common formulaic wish for longevity. While the four stanzas differ in some of the details, they are nothing more than variations on one another. If there were another three stanzas, they would do just the same; if a stanza went missing, nothing would be lost. In short, the entire poem can be seen as four different takes on a single repertoire, and stanza one, for reasons we do not know, ended up both in the beginning of the Mao shi poem and provided the gist of the text for the versions we now find in manuscripts. This does not privilege the Mao shi as “original”; it may simply mean that the formulas of the first stanza were the ones everyone knew. Thus, to quote a couplet from stanza one was, in effect, to quote what the entire poem was all about. That the Mao shi version ended up with four stanzas instead of just two or three does not mean that this is how the poem was known. Instead, variations of the same kind were always possible as long as the same basic structure remained in place.

From quotation patterns such as those for “Shi jiu,” and from the actual examples of the “Xi shuai” and “Jing zhi” poems, we can thus develop an idea of the constraints under which the composition of individual poems from larger poetic repertoires operated. Let us now review further examples within the received literature in order to reduce our reliance on looted manuscripts of uncertain authenticity. First, consider the case of the “Martiality” (“Wu” 大武) dance suite as it is described in a narrative Zuo zhuan 左傳 entry nominally dated to 597 b.c.e. In the following passage, I identify the corresponding poems in the Mao shi:

Without question, “Martiality” (or “Great Martiality” [“Da wu” 大武], as it is called elsewhere) was part of Eastern Zhou cultural memory and as such was repeatedly performed to reenact mimetically the mid-eleventh century b.c.e. Zhou conquest of the Shang. In the “Records of Music” (“Yueji” 樂記) chapter of the Records of Ritual (Liji 禮記), Confucius himself is quoted as saying “So did the Way of Zhou reach throughout the four quarters, as ritual and music were communicated and penetrated the realm—is it not appropriate that ‘Martiality’ was performed in slow, long-lasting moves?” (若此，則周道四達，禮樂交通。則夫《武》之遟久，不亦宜乎).Footnote ⁴⁹ Moreover, the dance suite was itself part of a larger performance repertoire that also included various versions of King Wu's 武 battle speech(es) or “harangue(s)” (shi 誓) that could be called upon when so desired. In the following Zuo zhuan account for the year 538 b.c.e.,Footnote ⁵⁰ the Chu minister Jiao Ju 椒舉 advised his ruler, King Ling of Chu 楚靈王 (r. 540–529 b.c.e.), on how to host the regional lords:

What we have here for both King Wu's harangue(s) and the dance suite attributed to him is not a privileging of the written record—there is no mention at all of any writings—but, instead, an account of a performance culture where existing repertoires of earlier performances and texts could be selectively reenacted. This, of course, is how ritual and cultural memory worked in premodern societies—and continue to work in contemporary religion—where defining moments of a society's past were and still are today continuously reimagined, reperformed, communicated, and perpetuated.Footnote ⁵² If genuine, the Qi ye and Zhou gong zhi qinwu manuscripts themselves were part of this culture of memory where practices of performance and writing constantly interacted and evolved together over time.

Most interestingly, the Zuo zhuan passage on King Wu's compositions mentions first a “eulogy” (song 頌) and then in addition “Martiality”; altogether, it quotes four poems from the “Eulogies of Zhou,” three of them as “stanzas” of “Martiality.” However, in the anthology of the Poetry, these quotations are not in stanzas of a single piece titled “Martiality” (which there is a poem of just twenty-eight characters in seven lines); instead, they all belong to different poems, each one with its own title. Considering that the Zuo zhuan passage mentions both a “sixth” and a “final” stanza, the suite must have contained at least seven stanzas altogether. For more than a century, modern scholars have wondered which poems from the “Eulogies” were included in the “Martiality” suite, with at least a dozen different proposals. But this Rankean quest of wie es eigentlich gewesen may itself be misguided: we do not know which songs were included and in what sequence they were performed. Instead, we should recognize that Zuo zhuan and Mao shi organize the texts of the “Eulogies of Zhou” in fundamentally different ways, and that this difference is not a problem to be solved but a phenomenon to be acknowledged. What if there never was a single, immutable, way of organizing “Martiality”? And what if the Mao shi arrangement of multiple short poems under individual titles was but a late textualization of a body of poetic material—the repertoire of “Martiality”—after its flexible, adaptive use in actual performances had already ceased? Once again, as in the case of “Jing zhi” and “Xi shuai,” a belief in individual authorship, in the moment of first composition, and in “the original poem” whence a line of written copies emerged, is the least plausible answer.

A wealth of other evidence shows that the combinatory and composite use of poetic repertoires in early China, and their use in ritual performance contexts, was the rule, not the exception. To stay with the “Eulogies of Zhou”: as I have noted elsewhere,Footnote ⁵³ most of the thirty-one “Eulogies” are very short: twenty-seven of them have just between eighteen and sixty characters each, and they are not discrete, self-contained textual units. Instead, some are closely interrelated, sharing entire lines or even couplets with one another but not with other poems, which marks them as expressions of particular repertoires.Footnote ⁵⁴ For example, of the mere thirty characters of “Year of Abundance” (“Feng nian” 豐年; Mao shi 279), sixteen are verbatim identical to verses in “Clear Away the Grass” (“Zai shan” 載芟; Mao shi 290), while “Clear Away the Grass” also shares three more lines with “Good Ploughs” (“Liang si” 良耜; Mao shi 291) and additional lines with four other neighboring texts.Footnote ⁵⁵ In other words, we are looking not at individual (let alone individually authored) songs but at “one poetry” (Owen) that was circumscribed by the formal and semantic constraints of a particular set of ritual demands.

Recently, Scott Cook in an excellent analysis of Zhou gong zhi qinwu has made similar observations regarding both “Jing zhi” and other “Eulogies of Zhou,” showing how lines and phrases from different poems appear to have been combined in different configurations according to different occasions in Western Zhou court ritual.Footnote ⁵⁶ In this, Cook develops further Fu Sinian's 傅斯年 (1896–1950) earlier discussionFootnote ⁵⁷ that the short “Eulogies of Zhou,” including “Jing zhi,” may once have been part of the repertoire of dance and music at the Zhou royal court. Writing in 1928, many decades before any of our recently discovered manuscripts had come to light, Fu perspicaciously noted the multiple connections between different poems within the “Eulogies of Zhou” and surmised that the “Eulogies” may have originally been multi-stanza compositions (or at least groups of poems that bore an inseparably close connection with one another) instead of the isolated short poems as which they appear in their Mao shi arrangement. He then assumed a history of loss and dissolution of the original forms of the “Eulogies of Zhou” due to their exclusive existence in the hands of Western Zhou court ritualists: once the Western Zhou court was gone, its music, dance, and poetry fell into disorder as well.

Cook, writing with the advantage of the manuscript evidence, advances a further hypothesis, namely, that the Mao shi preserves but selections of these earlier, more extensive suites that the Zhou court ritualists had employed in different configurations. This is entirely possible, but I am inclined to go just one step further. Instead of a history of loss and disintegration of the original poetic suites after the fall of the Western Zhou, or of a canonical selection of earlier texts in the Mao shi, I read the manuscript evidence as suggestive of a living poetic tradition, with the interaction of writing and performance in multiple, mutually independent strands continuing over many centuries beyond the Western Zhou, and possibly all the way through the Warring States. Moreover, this tradition included not just the “Eulogies of Zhou” but all sections and genres within the Poetry.

As suggested above for “Airs of the States” such as “Xi shuai” and “Shi jiu,” the notion of poetic repertoires is not limited to the “Eulogies of Zhou.” Furthermore, decades ago C. H. Wang proposed a “Weniad,” that is, an epic of King Wen that he reconstructed from a series of “Major Court Hymns” (“Daya” 大雅) in the Poetry.Footnote ⁵⁸ I do not agree with Wang's attempt to show that China, too, has some sort of an epic tradition,Footnote ⁵⁹ but his readings still reveal an important fact: the five poems he regrouped as the “Weniad” do in fact belong to a single narrative of the Zhou heroic past. Similar groupings on other topics can be found across both the “Major” and the “Minor Court Hymns” (“Xiaoya” 小雅).Footnote ⁶⁰ What these repertoires share are, once again, sets of topics and phrases that could be compiled in modular ways, yielding ever new poems that were always similar but never the same. This modular way of poetic composition extends further: large parts of the Lyrics from Chu (Chu ci 楚辭) anthology should be read as modular compositions from circumscribed repertoires—just consider the various cycles named as “Nine …” (jiu 九)Footnote ⁶¹—as should be Western Zhou bronze inscriptionsFootnote ⁶² and the stele inscriptions of the Qin First Emperor.Footnote ⁶³ The principle of modular composition evident in all such “one poetry” can also be observed in later Chinese poetry, most conspicuously in medieval gushi 古詩 and yuefu 樂府,Footnote ⁶⁴ and well beyond poetry even in traditional Chinese visual and material culture.Footnote ⁶⁵

Hypotheses or Conclusions?

Considering the phonological stability between Poetry quotations in manuscripts and their counterparts in the Mao shi,Footnote ⁶⁶ together with the fact that the manuscripts since c. 300 b.c.e. (Guodian, Shanghai Museum, Tsinghua University) include very few lines not found in the Mao shi, something close to the core of the Mao shi was known no later than in the fourth century b.c.e. While this corpus existed not yet in a fixed written form, it constituted the very center of elite education, as shown in texts like Zuo zhuan or in the various pronouncements by Confucius in the Analects: the Poetry was internalized through memorization and could be recited or written down whenever the situation called for it. And most importantly, as now in evidence from examples like “Xi shuai” and “Jing zhi,” even at a time when at least the signature lines of the Poetry had already solidified into quasi-canonical formulas, such textual performances in speech and writing still continued in mutually independent realizations of circumscribed repertoires.

What does all this mean for the formation, integrity, and transmission of our texts? How can the same material reappear in different configurations, without any ancient scholar ever being disturbed by this, or even just wondering about the original version of these materials? What even is an early Chinese text? In a culture where the same historical story, poem, or philosophical thought existed in multiple versions and could be attributed to multiple speakers, nobody before the founding of the empire in 221 b.c.e. ever claimed his version to be superior, older, or more authentic than any other one. In early China, there were arguments over ideas, but there was never an argument over texts.Footnote ⁶⁷

There are various ways of dealing with this situation, some better than others. One that the prevailing evidence will not support is the idea that we somehow have access to the original versions of ancient texts together with access to the minds of their authors, and that texts were stable in the way that we have come to think about texts today. These are mere fantasies.

Across all ancient cultures I am aware of, there is but one that privileged individual authorship from a very early moment:Footnote ⁶⁸ Greece, subsequently canonized and emulated in Rome and since the Renaissance hailed as the origin of, and model for, all “Western” civilization. In sixth- to fifth-century b.c.e. Greece, the following phenomena all happened within three generations, according to their reflection in the available sources: the Homeric epics became widely known, widely performed texts across a range of Greek cities; written texts began to circulate beyond their immediate contexts, and new texts were composed and read explicitly as writings; the words “poet” and “rhapsode” began to appear in written sources, denoting the “maker” and “performer” of epic poetry, respectively; and the name Homer appeared in the writings of Xenophanes (c. 570–c. 475 b.c.e.) and Heraclitus (c. 535–c. 475 b.c.e.) and was then quickly identified as the “maker” of poetry. Altogether, the paradigmatic cultural texts, the paradigmatic author, the widespread use of writing, the practice of reading, as well as the conceptual distinction between author and performer all emerged together and in relation to one another. Thus, when the Iliad and Odyssey became Pan-Hellenic texts, they simultaneously acquired their named author who, as a poet, “made” the texts in the way pottery was made (one of the earlier activities for which the verb poiéō was used—just as it was by Greek vase painters and sculptors since the sixth century as they signed their works).

As early as there were texts widely celebrated as literature, there were authors: not only Hesiod and Homer, but very soon also those who identified themselves at the outset of their works, such as Hecataeus (c. 550–c. 476 b.c.e.), Herodotus (c. 484–c. 425 b.c.e.), and Thucydides (c. 460–c. 395 b.c.e.). All three historians used the very first words of their books to introduce themselves and their goals. Far from introducing an element of questionable subjectivity, the historians named themselves at the outset to inspire trust: as history does not write nor explain itself, it becomes credible only through the credentials and explicitly presented methods of its author who takes personal responsibility for his text. Each author named his own local origin that placed him in a particular geographical, political, and cultural context, and from which he derived his authority and trustworthiness. Remarkably, the historians introduced themselves in the third person, not merely asserting but also historicizing themselves and their texts before turning to their first-person narrative histories. This practice must have developed in response to how the earlier poets—most notably Homer and Hesiod—had invoked not themselves but the muses (or the goddess, in the case of the Iliad) at the beginning of their texts: while for Homer and Hesiod, such invocation endowed the poetic text with authority because it was derived from divine inspiration, for Hecataeus, Herodotus, and Thucydides, the authority of their historical writing was rooted in a claim for truthfulness that was grounded in the historian's personal accomplishment and judicious use of existing sources. Moreover, these historians defined themselves as self-aware authors, situated themselves against the history they were writing, and insisted on the reliability of their information, as opposed to the myths and legends told by the earlier poets. Altogether, no later than in the second half of the fifth century b.c.e., the Greeks found it difficult to do without specific ideas about authors and the origins of their texts, be it in epic, dramatic, or historical writing.

Nothing like this existed in China before the Qin-Han imperial state, by which time Chinese written texts had been in existence for more than a millennium. Different civilizations develop different institutions and different impulses toward the stabilization of their textual traditions. In China, this process was overwhelmingly dominated by the imperial state since the late third century b.c.e., when a canon of classical texts was first monopolized at the Qin imperial court and competing versions that circulated beyond imperial control were censored and suppressed.Footnote ⁶⁹ Institutions such as the imperial academy, the imperial library, and the channels of written communication between the imperially appointed (and salaried) scholars, officials, and ultimately the emperor himself all contributed to a new textual culture in need of a stable historical past complete with genealogies of discrete texts, intellectual lineages, and individual authors. Discussions over specific versions and interpretations of the classics were officially sponsored at the imperial court, and texts had to be collated, defined, and created in newly standardized editions for the purposes of the imperial library as the repository of officially sanctioned knowledge.Footnote ⁷⁰ Repeatedly, the process of textual stabilization was one of reducing larger bodies of textual materials to smaller, more manageable ones, whether in poetry, historiography, or philosophy.Footnote ⁷¹

When thinking about ancient texts, we must not project either the paradigms of fifth-century Greece or the practices of the Chinese imperial state onto the textual culture of pre-imperial China. Instead, we are best advised—especially in light of the evidence from manuscripts—to let go of what earlier generations have taken as certainties, including assumptions of stable, discrete texts, their authors, and their linear transmission as writings. We know that Chinese texts before the empire could be written down repeatedly, and locally, but without any central control over their integrity and without concern over their origin or authorship. No philosophical, literary, historical, or technical manuscript from early China contains the name of its author. No pre-imperial manuscript excavated or looted that has a counterpart in the received tradition is in its internal structure identical to the transmitted version. Texts continued to change, and not always for the better.Footnote ⁷² For China before the empire, we are lacking any central institution—like the Temple in Jerusalem for the Hebrew BibleFootnote ⁷³ or the Ptolemaic Library at AlexandriaFootnote ⁷⁴—where a text was written, edited, and guarded, and from where it was then faithfully transmitted by mechanical visual copying.

It is fruitless to consider any newly discovered manuscript text as “better” or “more original” or “more reliable” than its received counterpart. Just because we now have one or two such manuscripts does not privilege any of them in any way: there could be a hundred more that are waiting to be discovered, not counting all those that no longer exist. Even worse are attempts to assimilate the rich textual and conceptual differences of newly discovered manuscripts to the traditional framework, instead of allowing these differences to challenge whatever we thought we already knew. The vast majority of manuscript texts are unknown from the received tradition; they are there not merely “to complement” (buchong 補充) the tradition, but to allow us to rethink the early development of Chinese textuality, including our all-too-familiar concepts and classifications that are entirely the result of early imperial attempts to retrospectively create a textual order in the service of the imperial state, its officials, and its scholars.

Texts before the empire were not just written differently; they were different texts. Every early text we know went through multiple stages of transmission, often both oral and written, and was reconstituted repeatedly in different forms along the way. In performance, textual repertoires were mobilized in response to situational needs, leading to a multiplicity of individual textual realizations none of which should be reified as a discrete, autonomous work. In writing, we know of different local and regional scribal traditions, of regional calligraphic systems, and of the fact that many ancient characters were eliminated when the Chinese script was standardized beginning with the Qin and Han empires. Each case of rewriting a text into a new script form involved acts of interpretation. Every transcription required a new decision on which graph to choose, opening some possibilities of meaning and interpretation while excluding others. From the macro-level of the integrity of an entire text to the intermediate level of its internal structure and further to the micro-level of individual character choices, early Chinese texts were reconstituted in new and different ways over and over again. Thus, early imperial commentaries were also not just attached to fixed, preexisting canonical texts; they created these texts, prominently among them the Poetry, anew by way of particular hermeneutic decisions.

Instead of regarding pre-imperial texts as stable objects, we are better advised to look for the processes of change and fluidity that may inform us about textual practices and the communities where such rich and varied practices were pursued. In this, we have to learn how to embrace and cherish uncertainty, and how to offer hypotheses, not conclusions. Given the extremely fragmentary nature of the evidence, the best we can do is formulate hypotheses that (a) can accommodate all the available evidence and, equally important, (b) are well informed by comparative perspectives. And while the next new manuscript is certain to appear, much remains to be gained from radically rethinking the received corpus and our traditional assumptions about it.