About this text: The Great Learning of the Pure and Real

1. Witnesses

The date of B

The title page reads, in the top band, 乾隆五十九年重鐫 "re-cut in the 59th year of Qianlong", i.e. 1794; the right column 真回老人著; the centre 清真大學. The publisher's line at lower left is worn and overwritten by a pen mark: "…齋藏板" (or 板/版) can be made out, the first two characters not safely (AI and eye readings: 敬?□齋). There is no preface and no colophon; the text ends 清真大學終 (p. 145). So the archive.org title's "Original 1648" does not hold for this copy: it is a Qing re-cutting of 1794 ("重鐫" implies an earlier cutting, which it does not name). The upload gives no source for "1648"; the date of the first edition was not checked here. The HathiTrust record itself was not reachable (bot check); the PDF's metadata say "Qingzhen da xue / Wang, Daiyu / HathiTrust".

2. Base and why

N: clear type, public domain, punctuated. Its head-notes, set above a wavy rule at the top of the page, are not the 1934 editor's, as was assumed: the same notes stand above the frame in B (1794), e.g. 開章一句，發諸未有 (N p. 11, B p. 5), 反復叮嚀，惟恐異見 (N p. 39, B p. 107). They are still left out (they are commentary, not Wang's text); the modern prefaces and the 1934 appendix 穆聖之遺教 (p. 40, PDF p. 50) are not reproduced; only Wang Daiyu's text. B supplied what library stamps cover in N in the opening (§1) and the closing verses (§31). The Internet Archive's text layer of B is unusable (OCR noise), and there is no text layer for N. B is not taken as base: it is a worn Google scan, with its own slips, and N is the text as read in the 20th century; B's readings are recorded in the apparatus, not adopted.

3. Collation

• Sections 1–31 against B, whole book (2026-09-28), with scripts/chinese/collate_qzdx.py. B's pages rendered at 300 dpi with pdftoppm, cropped to the frame and scaled to 1568 px. For each of the 71 half-leaves, headless Claude (opus, no tools) made two independent passes: a blind transcription of the page (no text given), and a compare pass (image plus our text for that stretch, located by aligning the blind transcription), listing every difference. A difference found by both passes was accepted as B's reading (200); one found by only one pass went to a third, A/B check on the image with the two readings in shuffled order, and, if not a mere variant form, to a fourth, open re-reading of enlarged quarter-pages; readings that failed were dropped. Every reading of B that was not a routine variant form was then checked against N itself (the 1934 page images, 300 dpi, same A/B method), to catch errors in our transcription; those N-check hits were looked at by eye at 600 dpi.
• Result: 278 entries in scripts/chinese/src/apparatus-qingzhen-daxue.json under collation (witness B, page = B's PDF page, type, and errorIn for slips), shown under each section in the reader. 71 substantive, 195 orthographic, 12 slips (8 in B, 4 in N). Sections 7–31 (never collated before): 60 substantive, 134 orthographic, 11 slips; sections 1–6: 11, 61, 1.
• Not listed: punctuation, 眞/真 and 爲/為, B's honorific raising, and the chapter headings (本然第一章, 真一本題 etc.), which are the same in both prints and are the passage headings here. Routine variant forms of B: 証 for 證, 于 for 於, 卽 for 即, 托 for 託, 单 for 單, 叅 for 參, 巳/已 for 己, 葢 for 蓋, 玆/兹 for 茲, 凢 for 凡, and 獨 cut as 蜀 beside 犬 (the AI read it as 歟, 獸 and the like).
• Taboo forms: B writes 玄 (元機 → 玄機, 妙元 → 妙玄) and 弘 (宏道 → 弘道) where N has the Qing substitutes 元 and 宏; N's 元首 for B's 君 (§19) may be a Republican change.
• Sections of B's scan: §1 pp. 5–11, §2–6 pp. 11–37, 真一 (§7–10) pp. 37–55, 數一 (§11–14) pp. 55–67, 體一 (§15–18) pp. 67–91, 結論 (§19–31) pp. 91–145.

• Notable readings of B (not adopted): §2 B lacks 各國皆治而後天下乃平，天下不平由於國有不治 (an eye-skip); §4 B has 穆罕默德 with the gloss ［譯曰聖諱］ where N has 穆海默德; §6, §15 B 數本之一 for 數一之一 (three times); §9 B 但靜則如如不動 for N's 但靜則動，動如不動; §10 B names the second of the six epithets 四海元首 (N 一切之主); §18 B 無方所 (N 無力所); §19 B heads the conclusion 總論 with its summary in small type (N 結論); §20 B 即如美心設玩，本爲驗人之觀察耳 … 設玩 against N's 即有善意設備，亦為天使用作驗人之觀察耳 … 設備; §21 B 粗心 (N 蠡心); §25 B 天闕 (N 天園); §29 B 房廊 (N 房廓), 掃除 (N 掃出).
• Before this: sections 1–6 were read from N with B consulted where the stamp covers N (§1); sections 7–31 from N alone, twice, at page and half-page size (34 doubtful characters in uncertain).
• N raises honorific terms (真一, 數一, 真主, 主僕, 獨一) to the head of a new column, as B does; the raising is not reproduced. N's ○ is given as 。.

4. Decisions (substantive)

Kept as printed and flagged as misprints of N, with B's reading beside them (resolved 2026-09-28 against B): 無力所 (§18: B 無方所), 蠡心 (§21: B 粗心), 但靜則動，動如不動 (§9: B 但靜則如如不動, the clean parallel to 動則紛紛不已; N's text is probably a corruption of 如如). Kept and confirmed by B: 闌 in 生死皆緣闌獨尊 (§22; B has the same glyph, 門 over 柬; 關 would make better sense in both). Also kept: 婆陀 (Buddha); variant forms 后, 岐, 悞, 勅, 卻.

Corrections to our text (errors of our transcription against N, found through B and checked on N at 600 dpi)

The AI English was adjusted to match (§3 "the Lord", §10 "decreed", §21 "read over", §25 "the harm extends", §31 "in root unlike").

5. Quotations

§4 quotes the Scripture (經云) and the Sage (聖云) without references; the Quranic allusions (Q 3:18; Q 33:72, §3) are identified by AI and marked. Sections 7–31 quote the Scripture and poems without references, and twice the Laozi (彼所謂「衆妙之門」, 「有名萬物之母」, §12, §14), which Wang sets to his own purpose.

6–8. Characters, IDs, analysis

As for the Tianfang xingli (see its record): Unihan, cjk-decomp, Kangxi via Wikisource; characters addressed <section>.<n>.

9. Translation

English by AI (Claude), marked, section by section. It was made without Murata's translation; her English (2000) was then read privately beside it, for sense only (see §12).

10. Term mappings

New for these chapters (scripts/chinese/src/terms.json): 本分 / 本為 (with 本然: dhāt, ṣifāt, afʿāl), 元勳 (Muḥammad), 代書 (the Pen, the First Intellect), 知認 / 見認 / 續認 (the knowledge, eye and truth of certainty), each after Murata 2000, pp. 75–76 (cited, not reproduced); the three transliterations with Wang's own glosses, 討黑德 = tawḥīd (習一), 以體哈德 = ittiḥād (自一), 穵哈德特 = waḥdat (純一) (source); 原有, 能有, 大筆, 前定, 普慈 / 獨慈, 初盟, 天園 (AI, with confidence); 無極 and 太極 as Wang places them (source, no Arabic).

11. Rights

N and B: public domain texts. The 1934 editorial matter is left out on purpose. Murata 2000 is in copyright: page citations only.

12. Open questions / to do

• Collated against B (done 2026-09-28, see §3). Still open: the publisher's name on B's title page (worn: …齋藏板); the HathiTrust record (bot check); a human check of the 71 substantive readings of B, which are AI readings of a worn scan (confidence is recorded per entry; §20's long divergence and §21 獨慈興/與 are the least certain).
• The first edition and other Qing printings are not on disk; a third witness would settle where N and B divide.
• Murata's translation (2000) was read privately beside the AI English, section by section, and every difference in sense decided from the Chinese (private notes, with page references: (private copy)). Of 77 differences, 7 were errors in our English and are corrected (§5 受造行造; §9 本無一物與同, "nothing was with Him", and 前定 as qadar; §16; §21; §22, where 能有 is the middle term of the triad; §31); 21 imprecisions were noted, of which the term renderings are now consistent (能有 Capacity-Being, 代理 the Deputy, 代書 the Deputy Scribe); 36 are places where the published translation misparses the Chinese, and 13 are open.
• Recheck against the scan: §7 遂已/遂己, §21 於己/於已. (§10 一切之主: N prints so, B has 四海元首; §9 靜則動: printed so in N, B 如如.)

13. Change log

• 2026-09-28: collated with B, the 1794 woodblock (identified from its title page; not "1648"), whole book: 278 variants recorded; 8 corrections of our transcription against N; the §31 stamp, 無力所, 蠡心, 靜則動 resolved; the head-notes and the §19 summary shown to be in the 1794 print already.
• 2026-09-28: 題綱 統說 in six sections, AI English, terms.
• 2026-09-28: the rest of the book, 真一, 數一, 體一, 結論 (sections 7–31, 6,417 characters), with the apparatus of uncertain readings and 18 new term mappings.