>The Rejected Pull Requests

A reject queue that just closes the PR teaches no one. This one left a written review on all 319 — and the reviews are the most honest part of the catalog.

Every library has a policy for what it admits. The Siku Quanshu Zongmu Tiyao had two tiers. The first — the main catalog — listed the 166 Yijing commentaries judged worth hand-copying into the imperial library. The second — the cunmu (存目, "preserved titles") — listed everything else they reviewed and decided was not worth the scribes' labor. For the Yijing section, that second tier runs to 319 entries across four juan.

The cunmu is usually described as a list of rejected books. That understates it. A list rejects silently. The cunmu rejects with commentary. For each of the 319, the compilers wrote a review — sometimes two sentences, sometimes a full paragraph — explaining precisely why the text didn't merit inclusion. The distinction mattered: merged texts were physically copied into all seven manuscript sets and distributed across the empire, becoming permanent canon. Closed texts got only a review — the one form of immortality the compilers were willing to grant.

Read across the 319 and the reasons cluster, though there was no formal rubric — each review is continuous prose.

Fraudulent attribution. The most entertaining class. A text claims an ancient author — Confucius's disciple Bu Zixia, the Wei sage Guan Lang, the Song master Shao Yong — but the compilers detect anachronisms that expose the forgery: vocabulary from the wrong period, institutional terms that didn't yet exist, citations of unwritten books. They are forensic about it, checking dynastic bibliographies and comparing prose style.

Derivative scholarship. The largest class by volume, the least colorful treatment. A commentary that restates positions earlier scholars expressed better, adding nothing. A typical verdict notes the author "mostly follows Cheng Yi and Zhu Xi" (大抵本程朱之說) and "does not go beyond what its predecessors already said" (不出先儒範圍). Two sentences. Next book.

Heterodox interpretation. Rejected not for poor scholarship but for an illegitimate framework — commentaries importing Daoist alchemy, Buddhist meditation theory, or folk religion into the Changes, treated as contaminations. The compilers don't object to those traditions existing — they catalog them elsewhere without hostility — but insist the Changes is a Confucian classic to be read within that tradition.

Poor craftsmanship. Sometimes it's just quality. A defensible position executed badly: muddled argument, careless textual work, confusion about what the original says. One author "attempted to reconcile the image-number and meaning-principle schools but succeeded only in confusing both" (欲兼取象數義理而兩無所據).

The most spectacular entries are the forgery exposures, approached with scholarly precision and barely concealed satisfaction. The Zixia Yizhuan (子夏易傳) takes the prize — attributed to Bu Zixia, claiming a lineage back to Confucius. The verdict: 易傳之名最古，其偽亦最古。而偽中又有偽焉，殆無盡矣 — "none claims greater antiquity than this book; yet none has generated forgery upon forgery, layer upon layer without end."

The original was already recognized as fabricated in the Tang. Then someone in the Song produced a new version claiming to be the real one — probably by the Tang-era Zhang Hu (張弧). But the text circulating in the compilers' own time matched neither: 然則今本又出偽託，不但非子夏書，亦並非張弧書矣 — "the current text is yet another fabrication; not only is it not Bu Zixia's book, it is not even Zhang Hu's." A forgery of a forgery of a forgery. The Guanshi Yizhuan (關氏易傳) fell to a simpler test: not listed in the Sui or Tang bibliographies (隋唐經籍志皆不著錄) — a damning absence for a supposed Northern Wei text.

The project ran inside a political environment, and some cunmu decisions reflect ideology. Late-Ming commentaries tied to the Donglin Academy (東林書院) and its activism drew harsher treatment than their scholarship alone warranted — when a Ming loyalist used the Changes to encode criticism of the order, the compilers had every reason to find the work "derivative" or "confused." The court's preference for evidential scholarship (考證學) consistently penalized the Wang Yangming (王陽明) tradition, which "abandoned the text and spoke only from the mind" (離經而空談心學). And the three imperial-sponsored Yijing works that open juan 6 — by the Kangxi, Yongzheng, and Qianlong emperors — all receive enthusiastic praise, while no private scholar's work gets comparable treatment. Reviewing your employer's own intellectual contributions while on his payroll introduces a certain bias.

The compilers were brilliant but not infallible. Later scholarship shows some cunmu judgments wrong — sometimes badly. Texts dismissed as derivative proved to preserve readings found nowhere else, gathered from local traditions the Beijing-based team never encountered. More substantially, their preference for the meaning-principle tradition led them to undervalue image-number works later found important: the Song hexagram-diagram tradition they treated as a dead end has proved relevant to the history of mathematics, and Shao Yong's binary-like ordering attracted Leibniz in the seventeenth century. A bibliography compiled under a specific set of assumptions reflects those assumptions — a legitimate set of criteria, but not the only possible one.

The most revealing thing about the cunmu is that it exists at all. The compilers didn't have to review the rejected texts — they could have simply excluded them and said nothing. Instead they wrote 319 reviews. They state the principle in their review of Yang Jian's commentary, a work so radical Zhu Xi said it "could be burned" (楊敬仲文字可毀): they preserve it anyway, citing Zhu Xi's own precedent — 存之正所以廢之, "preserving them is precisely what exposes them."

A forgery ignored will resurface; a derivative commentary merely excluded will be rediscovered by someone who doesn't know it's derivative. By putting on record exactly what's wrong, the compilers spare the next generation from making the same discovery independently. The cunmu is not a list of failures — it's an inoculation against future error.

The 319 rejected texts are as valuable as the 166 accepted ones: a negative map of the tradition, marking every interpretive dead end. Not censorship. Quality control with a paper trail.