How Well Did the Commentators Reverse-Engineer the Zhouyi? A Pre-registered Test of Two Interpretive Rules Against the Received Text

The line statements of the Zhouyi were compiled in the Western Zhou. The interpretive apparatus most often used to read them — centrality (zhong), correctness (zheng), correspondence (ying), and the adjacency relations cheng (supporting) and cheng (riding) — is set out in the Ten Wings, five centuries later. That apparatus therefore cannot have generated the line statements. It is a reverse-engineering of a text that already existed. This paper asks how good that reverse-engineering was, and treats the question as measurable rather than interpretive. We pre-registered two rules and hashed all 384 predictions before obtaining any line-statement text. Rule A uses only the twelve-character positional maxim of the Xici: the second line often brings praise, the third often misfortune, the fourth often apprehension, the fifth often merit. Rule B uses the full zhong-zheng-ying-bi apparatus. Judgments were assigned to the 384 line statements by a published deterministic keyword rule, not by human reading. On the 254 line statements that carry an explicit favourable or unfavourable judgment, Rule A is correct 66.9% of the time and Rule B 65.4%, against a majority-class baseline of 59.1%. On the 87 items both rules judge, Rule B is nominally higher (75.9% against 72.4%) but the difference is not significant (McNemar chi-square = 1.33). By the pre-registered criterion, the fuller apparatus shows no explanatory power independent of the positional maxim. We report the number, the corpus, the annotation rule, and a construct-validity problem that our own pre-registration failed to anticipate. An appendix records the formulations we tried and rejected, since each corresponds to a question a reader is likely to ask.