Principles of Philological Method

Chris Fraser

I. General Principles

1. General Aims and Criteria of Adequacy. Textual emendation and interpretation should be regarded as interrelated parts of a joint philological and interpretive theory that aims to explain all aspects of a text, including its meaning and structure and the condition of the received text.

a. Neither emendation nor interpretation has a more fundamental justificatory status. Rather, satisfactory hypotheses concerning emendation and interpretation must jointly provide the best explanation of the received text. Hence philological issues cannot be settled prior to or independently of interpretive issues and vice versa.

b. Such a joint philological-interpretive theory is justified to the extent that it provides the best explanation of a text.

c. The “best” explanation is one that scores higher overall on commonly accepted criteria of adequacy for a scientific hypothesis or theory than any alternative explanation does. These criteria include:

i. Testability: The hypothesis or theory entails correct consequences concerning observations other than those it was introduced to explain, and thus is not ad hoc.

1. For example, a proposed emendation might lead one to expect graphic errors or variants of a kind that are in fact found in other passages or texts.

ii. Fruitfulness or predictive power: The hypothesis tends to yield or contribute to correct explanations or predictions of other points not anticipated when it was introduced.

iii. Scope or comprehensiveness: The hypothesis explains more different types of observations than alternative views.

iv. Simplicity: The hypothesis is simpler than its rivals, largely, though not only, in making fewer assumptions.

1. For example, in interpretation, a simpler hypothesis is one that depicts the object of interpretation as conforming to familiar patterns of reasoning, such as avoiding obvious, inexplicable inconsistency or other errors in reasoning or perception. In textual emendation, a simpler hypothesis is one that offers a plausible explanation of the source of the graphic errors it posits.

v. Conservatism: The hypothesis is consistent with other well-justified beliefs, including beliefs about the text, its meaning, the context in which it was produced, and the history of its transmission.

1. For example, a proposed graphic emendation is more plausible when examples can be adduced of similar instances of graphic error or variation or when it yields an interpretation coherent with the most plausible interpretation of the immediate context of the emended passage.

d. In philological and interpretive research, the available evidence—the received text(s)—is limited and generally does not allow controlled experiments in which phenomena can be produced and observed specifically for the purpose of testing a hypothesis. Hence testability and fruitfulness play a smaller role than they do in experimental science. Still, these criteria remain relevant.

i. For instance, we might conclude that an explanation of one part of a text is not testable if it has no consequences at all concerning other parts of the text—that is, no other possible observation about the text could be inconsistent with the hypothesis.

ii. A fruitful explanation is one with consequences that agree well with further observations about the text. An explanation is not fruitful if its consequences are inconsistent with, or at least do not seem supported by, other aspects of the text.

1. For instance, the hypothesis that 文 is an error for the graphically similar 之 is fruitful because it explains not only one passage, but several.

2. By contrast, the hypothesis that Canon B66 alludes to a concept of “essence” suggests that we should find such a concept playing a role in other passages on semantics, such as A78, but we do not. Hence this hypothesis is not fruitful.

iii. As in the sciences, different hypotheses will score higher on different criteria, and a high score on one criterion (such as fruitfulness) may entail a lower score on another (such as conservativeness). Hence the judgment that one explanation is superior to another is always rough, imprecise, and conditional.

e. The criteria of adequacy for a theory allow us to define a concept of reasonable doubt: A hypothesis is subject to reasonable doubt when there are good reasons to believe that a contrary hypothesis provides an equally good or better explanation.

f. The criteria of adequacy and the notion of reasonable doubt allow us to define several possible degrees of justification of a hypothesis. Here these are formulated for hypotheses concerning textual emendation:

i. Strong: An emendation is strongly justified if the hypothesis that the emendation is the correct reading (i) is superior to alternative hypotheses, including the null hypothesis that the received text is correct, and (ii) is not subject to reasonable doubt.

ii. Moderate: An emendation is moderately justified if the hypothesis that the emendation is the correct reading (i) is superior to alternative hypotheses and (ii) is subject to reasonable doubt, but (iii) that doubt is significantly less justified than the hypothesis.

iii. Weak: An emendation is weakly justified if the hypothesis that the emendation is the correct reading (i) is superior to alternative hypotheses and (ii) is subject to reasonable doubt, and (iii) that doubt is only slightly less justified than the hypothesis.

iv. Speculative or unjustified: An emendation is speculative, or unjustified, if the hypothesis that the emendation is the correct reading is not superior to all alternative hypotheses, including the null hypothesis that the received text is correct. In such cases, the emendation should be rejected.

2. Textual Emendation. Textual emendation of any type is justified only to the extent that it contributes to the best explanation of how the text reached its present state. Such an explanation will include an account of the meaning of the text, the graphs originally used to express that meaning, and the processes by which the text became corrupt.

a. To be justified, an emendation must yield an interpretation of the text that explains the text better than interpretations supported by other emendations or by the received text.

i. If the resultant interpretation is only equal in explanatory power to another, competing interpretation of the received text, then emendation is unjustified.

ii. Any proposed interpretation must cohere with the best explanation of the overall meaning of the text.

b. A proposed emendation must be explicitly combined with a plausible explanation of the source of graphic error or other textual corruption. The combined explanation must cover both the meaning of the text and the existence of corruption.

i. This point is an application of the widely accepted canon of textual criticism that the best reading is that which best explains the others.

ii. Knowledge of human psychology and work habits suggests that scribes are more likely to err by modifying the content of a text from the unfamiliar and difficult to the familiar and easy than by doing the converse. Hence, other things being equal, by the criteria of simplicity and conservativeness, we should prefer emendations that attribute to the text errors of this type. (This point is a version of the traditional canon of textual criticism that the more difficult reading is to be preferred.)

iii. Examples of plausible sources of corruption are (i) copyists’ errors due to graphic similarity or similarity between adjacent phrases and (ii) lacunae due to damaged writing materials, such as broken bamboo strips or worn silk.

c. Since textual emendation is justified only if packaged together with an explanation of how the text became corrupt, other things being equal, explanations of the text that incorporate textual emendations are inherently more complex (or less simple) than explanations that accept the received text.

i. Hence, other things being equal, a theory that accepts the received text is to be preferred over one that rejects it. The text should be emended only if the proposed emendation is part of an explanation significantly better than an alternative that accepts the received text.

ii. If the received text is ungrammatical, unintelligible, or incoherent, a theory that accepts the received text must explain why an otherwise competent author would write a faulty passage. Typically, the complexity of such an explanation and the degree to which it conflicts with other justified beliefs about the author and the text will increase in direct proportion to the incomprehensibility of the text. In such cases, textual emendation will often yield an explanation that is simpler and more coherent.

d. The justification for all textual emendations should be stated explicitly in detail. The strengths and weaknesses of the evidence for proposed emendations and for alternative hypotheses should be clearly identified, and the degree of justification for each emendation should be explicitly labeled as “strong,” “moderate,” or “weak,” in accordance with the definitions of these terms in Section 1(f).

e. The hypothesis that a passage in the received text is faulty must be distinguished from hypotheses concerning how to emend the text. In some cases, the former hypothesis may be strongly justified, while no particular emendation is more than weakly justified.

i. Such cases should be identified and the degree of justification of the two hypotheses labeled accordingly.

ii. When no specific emendation or interpretation of a passage is at least weakly justified, none should be adopted. The passage should be identified as an open question for future research. Similarly, when an emendation is at best only weakly justified, that fact should be emphasized and the relevant passage identified as a topic for further research.

f. Conjectural emendation is inherently unlikely to be part of the best explanation of the text and may be attempted only when there is no reasonable doubt that the text is faulty, such as when the received text and all variants are ungrammatical or unintelligible.

i. For our purposes, “conjectural emendation” is defined as emendation that lacks support from variant readings, counterpart passages, contextual parallels or contrasts, or familiar or systematic instances of graphic error, graphic variation, or other patterns of textual corruption.

ii. Because it lacks such support and is by nature less simple than the null hypothesis that the received text is correct, a conjectural emendation can be part of the best explanation of the text only if there is no reasonable doubt that the null hypothesis explains the text very poorly, such as when the received text is unintelligible.

iii. Even in such cases, however, conjectural emendations are by nature unlikely to be part of the best explanation of the text, for, given the paucity of data, one or more alternative explanations are likely to score at least as high on the criteria of adequacy. Thus even if a conjectural emendation yields a reasonable interpretation of the text, it will rarely be justified, because it may still not be demonstrably superior to rival explanations.

iv. Hence emendations undertaken with no other evidential support than that, in the editor’s view, they render the meaning of the text more reasonable or intelligible are inherently unlikely to be justified.

1. This methodological point cannot be emphasized enough: The mere fact that a proposed emendation produces a reading that “makes sense” of a passage does not provide a reason to believe it is justified. For that, the emendation must be beyond reasonable doubt. See section 1(e).

II. Guidelines Concerning the Canons and Explanations

3. Textual divisions. The divisions between individual jīng 經 (“canons”) and shūo 說 (“explanations”) in the later Mohist texts are not marked in the received text. Some of the boundaries between sections are unclear because of syntactic ambiguity, textual corruption, or lacunae. Hence it is necessary to evaluate various hypotheses concerning how to divide the text into individual sections.

a. The main key to the division of the Canons and Explanations is the presence in the Explanations of “head characters,” graphs that now appear as part of the body of the text, but apparently were originally intended to serve as headings identifying the beginning of each shūo. The head character for each shūo consists of the first graph (or, in a few cases, pair of graphs) from the corresponding jīng. Hence the head characters provide an initial clue for identifying both the beginning of each shūo and the first graph of each jīng.

i. The present research adopts as a working hypothesis the heading hypothesis: Unless there are convincing grounds to the contrary in a particular case, we assume that every shūo has a heading and consider it a criterion of adequacy for a division hypothesis that it provide one for each shūo.

ii. To be justified, any hypothesis concerning the division of the text of the jīng shūo must cohere with the best explanation of the placement of the headings.

iii. In most cases, the presence of a head character and other contextual information points to a single, uncontroversial best explanation concerning textual divisions. However, in several dozen cases, corruption, ambiguity, and other factors render the correct textual divisions unclear and open to reasonable disagreement.

iv. The headings generally appear at the beginning of each shūo, but in a few places they appear to have been placed one or two graphs too early or too late. (They may originally have been written in the margin of the text, and thus later entered the text either above or below the first graph of the corresponding shūo.) Hence in some passages there may be good reason for transposing the head characters with one or more graphs preceding or following them.

v. Still, transposition cannot be accepted arbitrarily. It is justified only if (i) it yields a better interpretation of the text than interpretations available without emending the received text, and (ii) it is consistent with a plausible explanation of how the text reached its present form (how the graphs were originally misplaced).

b. Proposed divisions of the Canons and Explanations must yield grammatical sentences in classical Chinese and be consistent with the most plausible interpretation of the text.

i. Unintelligible, ungrammatical, or fragmented sentences at the beginning or end of a section in the received text of the Explanations seem best explained by the hypothesis that a heading has been misplaced.

ii. In such cases, textual division should be guided by the best interpretation of the passage and the best explanation of the cause of the misplacement.

1. For example, a hypothesis on which a heading was inserted several graphs out of place is prima facie less plausible than one on which it was inserted one graph out of place.

iii. Unintelligible, ungrammatical, or fragmented sentences at the beginning or end of a section in the received text of the Canons seem best explained by the hypothesis that the division between canons is mistaken or that textual damage has occurred. A plausible explanation must be given for any textual displacement or lacunae.

4. Graphic emendation. Graphic emendation refers to replacement of a graph in the received text by a different graph. In the Canons and Explanations, there are several common types of cases in which graphic emendation is likely to yield the best explanation of the text.

a. The received text is ungrammatical or unintelligible.

b. External evidence (the same passage in the earliest manuscript of the text in a distinct textual lineage) suggests a variant reading that yields a better interpretation of the text than the received version.

c. A discrepancy in terminology, content, or structure obtains between a jīng (canon) and the corresponding shūo (explanation), and one of the alternatives fits the context significantly better than the other.

i. The best explanation of the texts is that the shūo are intended to explicate the jīng and that the head characters of the shūo correspond to the first graph(s) of the relevant jīng. Hence graphic emendations that follow from these hypotheses are likely to be justified.

ii. If the first graph of a jīng is not identical to the head character of the corresponding shūo, at least one of the pair is overwhelmingly likely to be corrupt, and the graphs should be emended so that they are identical. Usually contextual information is sufficient to determine the best explanation of the discrepancy.

iii. If key terminology in the jīng and corresponding shūo is inconsistent, graphic error in one is likely. Again, usually contextual information is sufficient to determine the best explanation of the discrepancy.

iv. Apparent but defective parallelism between phrases in a jīng and its shūo may be best explained by graphic errors or other copyist’s errors in the transmission of the text.

d. Graphic permutation (confusion between similar graphs) is suspected. Emendation is especially likely to be justified in cases where a common, familiar graph appears to have been substituted for a relatively uncommon, technical usage.

i. To be justified, the emendation must correct a reading that seems nonsensical or unintelligible in the received text.

ii. The emended graph must be graphically similar in seal script (zhuan wen 篆文) to the graph in the received text.

iii. The emendation must contribute to the best interpretation of the passage, and the use of the emended graph must be supported by its use in thematically related passages elsewhere in the Mohist texts (or other closely related ancient texts).

e. A pattern of similar graphic error or permutation is found throughout the Canons and Explanations.

i. A repeated pattern of occurrence of a graph in contexts where a different, similar graph is more intelligible increases the plausibility of the hypothesis that permutation has occurred.

ii. Emendation is justified only if it yields an interpretation significantly better than that available from the received text. Hence emendation is likely to be justified only if the received text is unintelligible or nonsensical.

f. Emendation restores parallelism or contrast between different phrases in the same passage, which is needed to yield a reasonable interpretation of the text.

5. Transposition. Transposition of graphs, phrases, or sections of text may yield the best interpretation in cases such as the following.

a. The received text is unintelligible and transposition yields an intelligible, coherent reading.

b. External evidence provides a variant reading that yields a better interpretation of the text than the received version.

c. Transposition yields the best interpretation of both the transposed text and other passage(s) or graphs, such as those immediately preceding or following the transposed text in either the received or the emended version.

i. For example, the transposition restores a sentence that has been fragmented into two pieces.

d. The transposed text seems unintelligible in its location in the received version, but after transposition intelligibly fills a lacuna.

e. Transposition restores parallelism between a jīng and the corresponding shūo or between parts of a single passage.

f. Like other types of emendation, transposition should be supported by a plausible account of how the textual corruption in the received text occurred.

6. Deletion. Deletion, or excision of one or more graphs from the received text, may be justified in cases such as the following.

a. External evidence provides a variant reading that yields a better interpretation of the text than the received version.

b. Deletion yields an intelligible reading from an unintelligible passage in the received text.

c. Deletion restores parallelism or a contrast between phrases within a passage and thus yields the best interpretation of the text.

d. Deletion is supported by parallelism between a jīng and its shūo.

e. In general, copyists are unlikely to accidentally insert arbitrary graphs into a text. So deletion can be justified only in cases when a plausible explanation is available of why the purportedly extraneous graph(s) were interpolated. Such explanations may include:

i. Dittography, accidental repetition of a graph, phrase, or entire line of text.

ii. Accidental incorporation of marginal notes or sound glosses into the text.

iii. Intentional but mistaken incorporation of an entire bamboo strip into the text.

iv. Accidental or intentional incorporation of extraneous graphs that, together with the original text, form a word or phrase familiar to the scribe.

7. Interpolation. Interpolation, or inserting one or more graphs into the text to correct for haplography (accidental omission of graphs) or to restore lacunae, may be justified in cases such as the following.

a. External evidence provides a variant reading that yields a better interpretation of the text than the received version.

b. The text contains apparent but faulty or incomplete parallelism or contrast that is restored by the interpolation.

c. Apparent but faulty parallelism between a jīng and its shūo is restored by the interpolation.

d. Interpolation yielding an intelligible interpretation of the text will tend to be more justified to the extent that the interpolation is brief, the scribal error posited is easily explicable, and the received text is unintelligible.

e. Lacunae (lost portions of the text) can justifiably be restored only if external evidence or internal evidence such as parallels elsewhere in the text suggest content that provides the best explanation of the received text.

i. For instance, the restoration yields a compelling interpretation of the passages immediately preceding and following the lacunae, of the jīng corresponding to a missing shūo, or of the shūo corresponding to a missing jīng.

Appendix: Definitions

Conjectural emendation: Emendation based solely on a conjecture about what the original reading was, without support from variant readings, counterpart passages, textual parallels or contrasts, or familiar or systematic instances of graphic error, graphic variation, or other patterns of textual corruption. Because conjectural emendation lacks these sorts of support, it can legitimately be attempted only when there is no reasonable doubt that the text is faulty, such as when the received text and its variants are ungrammatical or unintelligible.

Contextual parallel or contrast: A parallel or contrasting phrase within the same section of a text. The structure of two purportedly parallel phrases may provide reasons for suspecting corruption in one of the two and emending accordingly.

Counterpart passage: A parallel or a counterpart passage in a different section of a text. The structure and content of two such passages may provide reasons for suspecting corruption in one of the two and emending accordingly, as when the content of a jīng is emended on the basis of the content of the corresponding jīng shūo.

Criteria of adequacy: See Section 1.

Degrees of justification: See Section 1.

Deletion: Excision of one or more graphs from the received text to correct for dittography or accidental interpolation of extraneous graphs.

Explanation, “best” or “better”: See Section 1.

External evidence: Evidence for emendation from outside the main edition of the received text, such as from other, independent manuscript lineages.

Graphic emendation: Replacement of a graph in the received text by a different graph, for which the graph in the received text is taken to be an error or variant.

Graphic error: A graph used erroneously in place of another, usually similar graph, as when 傳 might be used erroneously for 摶 or 攸 or 彼.

Graphic variant: A different graph used to write the same word as a familiar graph, as when 无 is used to write the same word as 無 wú (lacking) or 辯 is used to write 辨 biàn (distinction). Graphic variants may be common and familiar, as in these two examples, or nonstandard, as when 卞 is used for 辯 or 五 for 伍.

Haplography: Accidental omission of graphs from a text, typically due to homoioteleuton (“same ending”) or to homoioarcton (“same beginning”), when a scribe’s eye skips from the ending or beginning of one phrase to that of a similar phrase following it.

Interpolation: Inserting one or more graphs into the received text to correct for haplography (accidental omission of graphs) or lacunae (lost portions of the text).

Internal evidence: Evidence for emendation from within the received text.

Null hypothesis: The hypothesis that the received text is correct and no emendation is needed.

Permutation: Graphic error due to a scribe’s confusion of similar graphs.

Reasonable doubt: A hypothesis is subject to reasonable doubt when there are reasons to believe that a contrary hypothesis provides an equally good or better explanation.

Textual division: Division of the received text into distinct paragraphs or sections.

Textual emendation: Replacement, transposition, deletion, or insertion of graphs in the received text.

Transposition: Movement of one or more graphs to a position different from their position in the received text.

Variant reading: A difference in the content of what is purportedly the same passage in different exemplars of the received text.