title: Aligning Müller's Upaniṣads to the Sanskrit summary: How 851 verses of F. Max Müller's 1879–1884 English were attached to the Sanskrit of nine Upaniṣads — and why the rest were left blank. date: 2026-09-17
The source
Max Müller's The Upanishads (Sacred Books of the East, volumes 1 and 15) is public domain. English Wikisource holds proofread transcriptions of the scan pages for the Chāndogya, Kena, Īśa, Kaṭha, Muṇḍaka and Praśna; for the Taittirīya, Bṛhadāraṇyaka and Śvetāśvatara its pages are empty placeholders, so those were taken from the Internet Archive's OCR text of the same volume and are labelled scan text, unproofread wherever they appear.
The rule
A translation is attached to a verse only when both of these hold:
- Müller's section (khaṇḍa, vallī, brāhmaṇa…) has exactly as many numbered paragraphs as the Sanskrit section has verses.
- The evidence does not favour a shifted alignment. Müller keeps Sanskrit names and technical words in his English — udgîtha, prâna, Svetaketu — so each paragraph can be checked against the verse it is meant to translate and against its neighbours. A section where paragraphs point at neighbouring verses is refused whole.
For scan-text sources a third gate applies per verse: the paragraph must carry positive evidence for its own verse, or sit between two paragraphs that do.
Why the second rule exists
Bṛhadāraṇyaka 1.3 has 28 paragraphs in Müller and 28 verses in the Sanskrit. A count check passes. But Müller's "Lead me from the unreal to the real" sits in his paragraph 27, while the Sanskrit has it at verse 28. A count match alone would have attached the wrong English to the most quoted verse in the book. The evidence gate caught it, and the section is blank.
The result
| Text | Attached | Note |
|---|---|---|
| Chāndogya | 466 of 622 | ~20 sections differ in numbering between Müller's edition and the Sanskrit |
| Kaṭha | 97 of 119 | one vallī numbered differently |
| Praśna | 65 of 67 | |
| Muṇḍaka | 64 of 64 | created English-only; Sanskrit pending |
| Kena | 26 of 34 | khaṇḍa 1 numbered differently |
| Īśa | 18 of 18 | |
| Bṛhadāraṇyaka | 100 of 440 | scan text; per-verse gate |
| Śvetāśvatara | 13 of 113 | scan text; per-verse gate |
| Aitareya | 2 of 33 | Müller's paragraphing differs throughout |
The Taittirīya was refused entirely: only 21 of its 31 anuvākas parsed cleanly from the scan, and a text with a wrong structure is worse than no text.
What would improve it
A proofread transcription of volume 15 — which anyone can contribute to on Wikisource — would let the Bṛhadāraṇyaka, Śvetāśvatara and Taittirīya be aligned with the same confidence as the Chāndogya.