A critique of A.C. Graham's reconstruction of the "neo-Mohist canons.".

• Author: Geaney, Jane M.

The Neo-Mohist Canons are commonly thought to be the closest thing to logic in ancient China. A. C. Graham's reconstruction of this almost unintelligible text has been hailed as "the single most important study on Chinese logic ever published."(1) Graham suggests that the Canons also contain the germs of Chinese science, destined to be undeveloped due to poor preservation of this text. Sinologists have used Graham's reconstruction not only to understand Neo-Mohist logic and science, but also to elucidate methods of argumentation and technical terminology throughout ancient China. This article questions whether we can in fact rely on Graham's reconstruction. According to Graham, an "organizing principle must be identified if the items [in the Canons] are to be read in context."(2) But the organizing principle Graham selects for reconstructing the text is questionable. The organizing principle determines the order and the themes that provide the context for interpretation. If it is called into question, we lack the necessary context for interpreting the Canons. Given the countless questions about line-breaks and emendations of characters in the Canons, it is by no means easy to determine, on a case-by-case basis for each Canon, whether it is possible to reject Graham's organizing principle for the Canons as a whole but still to retain his translation (or even, in many cases, his decision about what counts as the beginning and ending of a "Canon" and an "Explanation"). Although Graham's reconstruction is a monumental achievement and we may have nothing better to use in its place, I fear that we do not yet have a reliable source of Neo-Mohist thought.

In what follows I shall first identify two general problems with Graham's proposed restructuring of the Canons: the fact that there are two apparently similar sections that Graham believes refer to distinctly different disciplines, and the fact that there are gaps in his construction of the Canons into supposedly parallel halves. Next, I shall question Graham's contention (fundamental to his restructuring of the Canons) that the Mohist divides the world into an eternally necessary realm and a transient, non-necessary realm. Graham contends that in the eternal realm knowledge is necessary, whereas in the transient realm procedures for knowledge are merely consistent, and that neither realm spawns epistemological questions. This separation of a necessary, atemporal realm from a non-necessary, temporal realm does not seem grounded in the thought of ancient China. And the intelligibility of Graham's reconstruction is not compelling enough to justify his theorizing that the Neo-Mohists invented such a world-view. If it is the case that the Mohist raises no epistemological questions, perhaps it is not because he believes knowledge is necessary (as Graham argues), but because, as elsewhere in ancient China, there is no drastic separation between realms such that skeptical questions might be raised concerning their reconnection.

Briefly, Graham's presentation of the history of the neo-Mohist Canons is as follows. The Canons were written sometime around 300 B.C. Unfortunately, when a complete text of the book called Mozi was assembled for the [TABULAR DATA OMITTED] Han imperial library (in the last century B.C.), the Canons (which make up pian 40 and 41 of the Mozi) had already been separated from their Explanations (which make up plan 42 and 43). Moreover, the sections Graham calls "Expounding the Canons" (which he abbreviates as EC) and "Names and Objects" (which he abbreviates as NO) had become fragments, respectively called "Bigger Pick" (pian 44) and "Smaller Pick" (pian 45), the lead characters (the first word of each Canon repeated as a heading) had been incorporated into the text. What Graham considers to be the text's five divisions (if they exist at all) were not marked. To make matters worse, before the end of the Sui dynasty (581-618), a text of the Mozi consisting of only the first thirteen chapters began to circulate. For about a thousand years thereafter no one read the complete text, although it survived in the Daoist Patrology. Eventually, the complete text was reprinted in what is known as the "Lu edition," based on the Sung Daoist Patrology (1552), and the "Tang edition," based on the Ming Daoist Patrology (1553). The first modern commentary on the Mozi was not written till the mid-1700s, and it was only in 1894 that Sun Yirang wrote his great commentary on the text.

Even after Sun Yirang's commentary, the Canons remained notoriously difficult to interpret. Graham proposes an outline that is intended to make the Canons intelligible. This outline splits the Canons in half, then splits each side into five parts in order to create a context from which to interpret each Canon. (Graham refers to these five parts as "sections" one through four, with the third section being a bridging sequence; see chart below.) The Canons were traditionally divided into "Parts" A and B. Graham calls this break "arbitrary" and divides the text at Canon A 88. Everything before that he refers to as "definitions" (A 1-87) and everything after as "propositions" (A 88-B 82). With his new arrangement, Graham believes that what he calls the "definitions" and the "propositions" can be seen to follow five more or less parallel sections with corresponding themes. This results in the model, supra, for the Canons.

This model makes it possible for Graham to circumscribe the topic of each of the Canons within one of the five sections. Accordingly, the fivefold order behind the Canons (a few pieces of which are actually absent from the Canons) is something like this:

1. Description = relating names to objects (transient).

2. Ethics = explaining how to act (transient). - Bridging sequence, on change.

3. Sciences = explaining objects (eternally necessary).

4. Disputation or Logic = explaining names (eternally necessary).

As a result of this order, Graham claims that, with the exception of the final section on logic, for every Canon there is enough of a context to establish "its general theme."(4)

However, in many ways these themes that are supposed to provide the context for interpretation are forced. Specifically, Graham himself admits that to us there appear to be two sections whose theme is logic (that is, in Graham's wording, "the realm of names"). But this, instead of causing him to question his theme-arrangement, prompts him to speculate that the Mohist "must be looking at logical problems from a different viewpoint which it would help us to locate."(5) Rather than search for a peculiarly Chinese approach to logic, we might instead reexamine Graham's division of the text and the parallel themes that lead him to such speculation.

In an attempt to distinguish between the two sections that both resemble "logic," Graham admits that the first section he identifies, which he calls "description" (names and objects(6)), "shares most of its terminology with the fourth discipline, disputation proper ['logic' or 'names only']."(7) In spite of this common terminology, Graham insists that there must be a difference between the two sections - that is, that the Mohists believe there is a difference between the "realm of names and objects" and the "realm of names." (Significantly, this difference involves the temporal nature of the two realms, as we shall see below.) Yet, Graham admits that this difference "emerges distinctly only in the two sequences of propositions" (i.e., in two of the sections on the side of what Graham calls the "propositions" half of the Canons). He notes that "the difference . . . is clear enough in the propositions but remains nearly invisible in the definitions."(8) However, he also maintains that in the propositions, where the distinction is supposed to "emerge distinctly," "the Mohist seems especially concerned with the fundamental terms which the first discipline shares with the rest, first of all the word ku 'reason'."(9) In other words, in the definitions half, Graham is unable to point to any difference between description and disputation. Moreover, in the propositions half, instead of the difference emerging distinctly," the shared terms emerge. Yet Graham does not permit this to invalidate his division of the two halves - the definitions and the propositions. Nor does it make him doubt what he claims to be the "parallelism of general themes" that seems to be his main justification for splitting the Canons just at that particular point, and referring to the two divisions as definitions and propositions. Instead, he relies solely on the following characterization of the first and fourth sections of the propositions for evidence that one of them is about names and objects (description) and the other is only about names (logic): he calls the first section "a close-knit sequence laying down procedures for deciding what is so of objects" and the last section "a series of miscellaneous propositions shown by logical ['names only' or 'disputation'] analysis to be admissible, self-refuting, consistent, unnecessary, inadmissible unless a condition is fulfilled" (emphasis added). Because what is at issue is precisely whether or not the first section is about "deciding what is so of objects" (as opposed to names only), this characterization of Graham's actually amounts to a claim that the first section is "a close-knit sequence" and the fourth is "a series of miscellaneous propositions."

There is no doubt that the fourth section is a miscellaneous series. In contrast to it, the first section may seem "close-knit." But that is unrelated to the question of whether the first section is about "description" (names and objects) or the fourth is about "logic" (names). The fourth section, which Graham says is about names, contains items B 52 (which appears to be about supporting weights from hairs) and B 62 (which has something to do with the relation of a spherical object and...