| Work / author / dates | Rasāʾil Ikhwān al-Ṣafāʾ wa-Khullān al-Wafāʾ (رسائل إخوان الصفاء وخلّان الوفاء), "The Epistles of the Brethren of Purity and the Friends of Loyalty": an encyclopaedia of 52 epistles in four sections (mathematical, natural-bodily, psychic-intellectual, nomic-theological) with a summary (al-Jāmiʿa). Anonymous; between the late 3rd/9th and the late 4th/10th century; authorship CONTESTED (§0). |
| Part added | A selection, one text in two parts (the author's scope, 4 Oct 2026): I animals: Epistle 22, the animals' case against man before the king of the jinn (faṣl 3–6 and the closing faṣl), ¶1–35; II principles: Epistle 32, the intellectual principles after the Pythagoreans (whole), ¶36–72, and the close of Epistle 13 with the definition of philosophy, ¶73–74. 74 passages, 9,187 words. Queued: Epistle 1 (number) and Epistle 5 (music), see docs/ROADMAP.md. |
| Reader route | /read/ikhwan (contents) · /read/ikhwan/animals · /read/ikhwan/principles · about page /read/ikhwan/about |
| Kind | Philosophical treatises (philosophy), §14 |
| Status | Hindawi/OpenITI text collated word by word against the Cairo 1928 print (vol. 3 with the Bombay lithograph), 7 corrections · AI-vowelled throughout (the print is bare) · AI English ✦ unreviewed · not yet checked by a person · reviewed by: — |
Structure. Part I (Epistle 22, «on how the animals were made and their kinds», of which the debate is the larger half): when Adam's children had spread over land and sea and yoked the beasts, a party of humans is cast up on an island in the Green Sea ruled by the jinn king Bīrāst (elsewhere Bīwarāsb) the Wise; the beasts complain of them, and the king hears both sides (¶1–6); the humans' spokesman claims lordship from the Quran, the beasts' spokesman answers that the verses speak of God's favours and not of ownership, and recalls that the beasts lived on the earth before Adam (¶7–12); the king summons his jinn and jurists (¶12–14); the humans boast of their upright form, the beasts answer that every form suits its use (¶15–17), then of their senses and reason (¶18–19); the humans claim lordship by buying and selling, feeding and housing them, and the pleas follow: donkey, ox, ram, camel, elephant, horse, mule, the pig (whom no one can classify) and the hare (¶20–29). The cut then passes over the long middle of the debate (the delegations of the beasts of prey, the birds, the insects, the sea creatures and the reptiles, the jinn sages' speeches, the human speakers of many nations) to the closing faṣl: the Ḥijāzī orator claims the resurrection and Paradise for humans alone; the nightingale answers that the beasts are also spared the punishment; the last claim, the saints («the friends of God») and the portrait of the perfect man, «Persian by descent, Arab in religion …»; the king's verdict that the animals be under the humans' command «until the cycle begins anew» (ḥattā yustaʾnafa al-dawr); and the Brethren's own word: do not take this epistle for children's play (¶30–35). Part II (Epistle 32): Pythagoras's doctrine that beings are ordered after the nature of number; the dyads, triads, tetrads; the one as the root of number as the Creator is of beings; then the ninefold order, each rank after the one before as each number after the last: Creator (one), Intellect (two), Soul (three), Matter (four), Nature (five), Body (six), the spheres (seven), the elements (eight), the generated beings (nine), with minerals, plants and animals as tens, hundreds and thousands (¶36–42); existence, duration, completeness and perfection (¶43); body, matter and form (¶44–47); how Intellect, Soul and Matter receive the Creator's effusion, and why matter, far from Him, desires nothing (¶48–50); questions on the principles, and a short poem (¶51); the spiritual and bodily principles in their ranks, from the active Intellect down through the spheres, the elements and the three kingdoms to the human being as a small world, and the soul's powers in minerals, plants, animals and the embryo (¶52–72). Part II, close (Epistle 13): logic as the balance and instrument of philosophy, because «it has been said in the definition of philosophy that it is becoming like God as far as is humanly possible» (al-tashabbuh bi-l-ilāh bi-ḥasab al-ṭāqa al-insāniyya), and the Brethren's gloss on «human capacity»: truth in speech, right belief, sound knowledge, good character, good acts, sound work and craft (¶73–74).
0. Authorship: CONTESTED
The authors present themselves only as an anonymous brotherhood. Graded CONTESTED: five positions, none with a consensus (C. Baffioni, SEP "Ikhwān al-Safāʾ", rev. 2025, §§1, 3). Sources checked by the companion sweep (app/src/content/companions/ikhwan.json → authorship):
- A Basran circle of the later 4th/10th century (the older view). Abū Ḥayyān al-Tawḥīdī, al-Imtāʿ wa-l-muʾānasa (Baghdad, c. 373/983), names Abū Sulaymān al-Bustī (al-Maqdisī), Abū al-Ḥasan al-Zanjānī, Abū Aḥmad al-Nahrajūrī (al-Mihrajānī), al-ʿAwfī and his friend Zayd b. Rifāʿa. P. E. Walker, Encyclopaedia Iranica VIII/3 (1998) pp. 294–297; S. M. Stern, Islamic Studies 3 (1964), repr. Studies in Early Ismāʿīlism (1983), DOI; J. L. Kraemer, Philosophy in the Renaissance of Islam (1986).
- An Ismaili work. The Ṭayyibī tradition gives them to the hidden imam Aḥmad b. ʿAbd Allāh (attested first in Idrīs ʿImād al-Dīn, d. 872/1468: D. De Smet, Shii Studies Review 1 (2017), DOI); Abbas Hamdani dates the corpus 873–909, before the Fatimids (IJMES 9 (1978), DOI; Arabica 26 (1979), DOI); Y. Marquet, Studia Islamica 61 (1985). Against: no Fatimid-era Ismaili text clearly cites them (Walker), and Ismaili use begins in the 6th/12th century (Stern, Diwald, De Smet, per the SEP).
- Shiʿi-adjacent and independent: a circle with Shiʿi sympathies and an «Ismaili tendency», not the official mission (Stern, Bausani, per the SEP); I. R. Netton, Muslim Neoplatonists (1982) and REP (1998); G. de Callataÿ, Ikhwan al-Safa' (2005).
- Earlier, and known in al-Andalus by c. 950 through Maslama b. Qāsim al-Qurṭubī (d. 353/964), which makes al-Tawḥīdī's men at most redactors: M. Fierro, Studia Islamica 84 (1996); G. de Callataÿ, Al-Qanṭara 34 (2013); W. Madelung (2018, per the SEP); G. de Vaulx d'Arcy, REMMM 146 (2019), who reads al-Tawḥīdī's «17th night» as a refutation.
- One author, al-Kindī's pupil Aḥmad b. al-Ṭayyib al-Sarakhsī (d. 286/899): de Vaulx d'Arcy (2016 thesis; Les Belles Lettres 2019), a minority view.
In the reader: the author line says «Ikhwān al-Ṣafāʾ (the Brethren of Purity)» and the dates «4th/10th century» (the Cairo print's own framing; the range runs from the late 3rd/9th). The text is filed under Philosophical treatises and not under Ismaili philosophy, since that would take position 2 as settled (§14). The Platonic Current's node keeps its «dating disputed». What our passages themselves show: the verdict's «until the cycle begins anew» (dawr), the «friends of God» whose portrait closes the debate, and «the imams, the trustees» (al-aʾimma wa-l-awṣiyāʾ) among the blessed in ¶32 are Shiʿi in colour; the colophon blesses «Muḥammad and his family, the guiding imams». None of this decides between positions 2 and 3.
1. Witnesses
| Siglum | Witness | Kind | Vowelled? | Licence | Location |
|---|---|---|---|---|---|
| C | Rasāʾil Ikhwān al-Ṣafāʾ wa-khullān al-wafāʾ, ed. Khayr al-Dīn al-Ziriklī, 4 vols (Cairo: al-Maktaba al-Tijāriyya al-Kubrā / al-Maṭbaʿa al-ʿArabiyya, 1347/1928), with Ṭāhā Ḥusayn's introduction and Aḥmad Zakī Pāshā's essay. Ep. 13's close vol. 1 p. 342; Ep. 22 vol. 2 pp. 152, 173–187, 314–317; Ep. 32 vol. 3 pp. 182–199 | standard print (no apparatus) | no | PD in the US (1928); al-Ziriklī d. 1976 | archive.org AAlexandrina-192979 (vols 1–2, 300 ppi grey), AAlexandrina-145864 (same, bitonal), 3_20250821_20250821_2209 (vol. 3, 72 ppi, degraded). Local: data/raw/arabic/ikhwan/ (ACQUISITIONS.md, SHA-256 per file) |
| H | Hindawi Foundation's digital text of C, via OpenITI 0375IkhwanSafa.Rasail.Hindawi095926405Vols-ara1 (editor named: al-Ziriklī) | digital | no | CC BY-NC-SA 4.0 | GitHub; every passage links its lines |
| B | Bombay lithograph, 4 parts, 1305–06/1887–89 (Maṭbaʿat Nukhbat al-Akhbār); Ep. 13 qism 1 pp. 125–130, Ep. 22 qism 2 pp. 119–245, Ep. 32 qism 3 pp. 2–16 | lithograph | partly | PD | DLI scans 323847, 323864, 323867 |
| D-ar | F. Dieterici (ed.), Thier und Mensch vor dem König der Genien, 2nd ed. (Leipzig 1881), Arabic, vocalised; his Abhandlungen … in Auswahl (1883–86) | edition | yes | PD | archive.org AAlexandrina-161831; dieabhandlungend01–03ikhw. Held, not collated |
| — | Manuscripts: Istanbul, Atıf Efendi 1681 (578/1182, the oldest dated), Esad Efendi 3638 (686/1287), not online; Paris, BnF Arabe 6647–6648 (675/1277) on Gallica, which volume holds our epistles not checked | MSS | known through the companion sweep | ||
| — | Critical editions (cited only): Goodman & McGregor, Epistle 22 (OUP/IIS 2009; IIS 2010); Walker, Poonawala, Simonowitz & de Callataÿ, Epistles 32–36 (OUP/IIS 2015); Baffioni, On Logic, Epistles 10–14 (2010). Beirut, Dār Ṣādir 1957 (in copyright) | copyrighted | not held |
Which edition H is. C: the vol. 1 title page («عني بتصحيحه خير الدين الزركلي», Ṭāhā Ḥusayn, Aḥmad Zakī Pāshā), the same division into four ajzāʾ and the same epistle titles, wording that matches line for line in every page collated, and shared quirks: the king «بيراست الحكيم لقبه شاه مردان» on the island «صاغون» (Dieterici's MS: Bīvarasp, Balāçāgūn), the order of the pleas starting with the donkey and the ox, and the conflated quotation «وتلك الأيام نداولها بين الناس وما يعقلها إلا العالمون» (Q 3:140 run into 29:43). H modernises spelling and punctuation and numbers the faṣls; no page breaks of C survive in it (they were found by matching, §7).
2. Base and why
H, corrected against C. C is the print H digitises and is public domain in the US; the Bombay lithograph is a different witness (its order of the mathematical epistles differs) and was used where C is illegible. The IIS/OUP critical editions are in copyright and not held. The other public-domain Arabic (Dieterici's) is held for a later collation of Part I.
3. Collation
Five AI readers, one per page range, compared H with the page images of C word by word (data/raw/arabic/ikhwan/work/collation/A–E.json; method in work/collation/INSTRUCTIONS.md): about 9,200 words, every passage. Vol. 3 of C exists only in a 72-ppi scan, so Epistle 32 was read in B and each disagreement checked against C's outline. 27 differences: 9 where C reads better (7 applied, 2 kept as doubtful), 6 where H corrects a misprint of C (kept: e.g. «تتخاص» for نتخلص, case endings), 12 variants kept as H has them. Applied (data/sources/ikhwan-corrections.json, logged in ikhwan-base.json → changes, type "collation"):
| ¶ | H | Adopted | Basis |
|---|---|---|---|
| 15 | الفصيلة | الفضيلة | C vol. 2 p. 180 (the dot clear in the bitonal scan) |
| 21 | فقال الحمار: أيها الملك | فقال: أيها الملك | C vol. 2 p. 183 (the second «الحمار» is only in H) |
| 31 | شجرة الزقزم | شجرة الزقوم | C vol. 2 p. 315 |
| 41 | ومشنؤه | ومنشؤه | C vol. 3 p. 185 (outline) and B p. 4 |
| 41 | بالواحد بعد كل عدد | بالواحد يعد كل عدد | B p. 4 and the sense; C's dots illegible |
| 65 | بالحجمة | بالمحجمة | C vol. 3 p. 196 (partly) and B p. 12 |
| 69 | ولا يغيضها | ولا يفيضها | B p. 14 and the sense (fayḍ); C's dots illegible |
4. Decisions and doubtful readings (kept as printed; ikhwan-base.json → uncertain)
- ¶52 «الرتبة الكلية»: the sense needs «النفس الكلية» (the next clause: «وانبجس من النفس»), which B has; C seems to repeat «الرتبة» and H put the words in «». Kept as printed; the AI English follows the sense with a bracketed note. A person should check C at better resolution.
- ¶64 «ثم دفعها لها بالقوة الهاضمة لتحيلها إلى ذاتها» stands twice in C (and H); B has it once. Kept.
- ¶67 B «واستؤنف» (C illegible); ¶63 B «صار» for «صارا»; ¶65 «ثقلت» (B «نقلت»; an earlier parallel suggests «فضلت»).
- ¶20 C repeats «اذا مرضت»; ¶21 C «لشفقة»; ¶25 C quotes Q 24:22 as «وليعفوا أو ليصفحوا» and Q 43:13–14 with «وتقولون … منقلبون»: our text has the Quran's wording, restored from Tanzil.
- ¶21 the conflated quotation of Q 3:140 + 29:43 is the print's own (and Dieterici's Arabic); kept.
5. Quotations verified
28 Quran quotations tagged in H (@QUR), each matched to its verse by consonantal skeleton in the Tanzil text and restored verbatim (scripts/recitals/restore_quran.py; 28/28): in Part I (18) Q 16:5–8, 23:22, 43:13–14, 22:37, 95:4, 82:8, 20:50, 24:22, 45:14, 6:38, 11:6, 39:73; in Part II (10) Q 27:59 (the opening doxology), 51:49, 24:21, 95:4–8, 21:104, 22:5. scripts/quran_refs.py adds automatic matches (stock formulas labelled). No ḥadīth is quoted in our passages.
6. Vowelling
C is unvowelled, so every mark is AI (scripts/recitals/prepare.py ikhwan, headless Claude Opus, letters verified unchanged; cached per passage under data/raw/cache/ikhwan/). Words with two defensible vowellings are listed per passage (ambiguous). Dieterici's vocalised Arabic of Part I (1881) is held and not yet compared. CAMeL disagrees with the AI vowelling on 349 words.
7. Segmentation and word IDs
74 passages of whole paragraphs (about 110–190 words), each heading opening a passage (scripts/recitals/ikhwan_base.py); headings for the epistles and the numbered faṣls (our English labels). Word IDs ik:<passage>:<word>. Print pages per passage from the collation (data/sources/ikhwan-pages.json, the H line in which each page of C begins). Paged reader: the base file's parts make the reader show one part a page (/read/ikhwan/<part>; lib/pagedProse.ts, components/LibraryReader.tsx); old links to /read/ikhwan#p-N or ?w=ik:N:… are sent on to the right part.
8. Analysis layer
CAMeL Tools 1.6 via scripts/fusus/analyze_ch.py ikhwan.json ikhwan-analysis.json ik: 9,129 words, 5,939 with a root, 908 distinct roots, 349 vowel conflicts.
9. Commentaries aligned
None exists. The Brethren's own al-Risāla al-Jāmiʿa restates their doctrine; not held.
10. Translation
AI (Claude Opus) from the vowelled Arabic, scripts/recitals/translate.py ikhwan opus, told not to reproduce or imitate the published translations and given none; glossary in data/sources/ikhwan-meta.json (tashabbuh bi-l-ilāh "becoming like God", al-ṭāqa al-insāniyya "human capacity", al-nafs al-kulliyya "the universal Soul", al-hayūlā al-ūlā "prime matter", fayḍ "effusion", zaʿīm "spokesman", arbāb/ʿabīd "masters/slaves", taskhīr "subjection" …). Word glosses and term notes from the same run. All ✦ unreviewed.
10b. Translation triangulation
scripts/triangulate.py ikhwan against four public-domain translations, each cut to our passages from the page images (by AI readers, checked word by word): Dieterici 1858 (Der Streit zwischen Mensch und Thier, pp. 1–23, 213–218; data/sources/ikhwan-tr-dieterici-1858.json), Dieterici 1868/1872 (Logik und Psychologie p. 59; Die Lehre von der Weltseele pp. 1–26; …-tr-dieterici-1872.json), and the two English versions of 1869 made from Ikrām ʿAlī's Urdu adaptation (1810), Platts and Dowson (…-tr-platts.json, …-tr-dowson.json; indirect witnesses, Part I only, ¶35 has none). Dieterici translates other manuscripts: in 1858 the king is Bīvarasp and the island Balāçāgūn, the verdict lacks «until the cycle begins anew» and the coda is fuller; his pp. 1–10 of 1872 render a longer, reordered recension of ¶36–42, while pp. 11–26 follow ¶43–72 closely. His definition of philosophy: «die Aehnlichwerdung Gottes so weit es der Mensch vermag» (1868 p. 59). The IIS/OUP translations are not held (copyright), so the triangulation could not use them. Result: 77 divergences, 22 judged serious by the model. Curated (scripts/recitals/ikhwan_differ.py): 11 demoted (not shown) because they come from the Urdu adaptation or Dieterici's other manuscript, not from two readings of our Arabic; 3 notes added where the letters of our text are in question: ¶2 «الحرس» guarding or «الحرث» ploughing (one dot; all translators plough), ¶27 «يغضبوننا» anger or «يبغضوننا» hate (one dot; the translators hate), ¶52 «الرتبة الكلية» or «النفس الكلية» (Cairo against Bombay and Dieterici). 14 shown, among them ¶17 «أعطى كل شيء خلقه» (Q 20:50: what God gave, and to whom), ¶20 «نعرض عنها», ¶21 the Rūm (Byzantines, or Turks in Dowson), ¶23 «عواري», ¶30 «النشور», ¶32 «مشفقون», ¶34 «الغرباء» (strangers, or the Urdu sense «poor»). AI-English flags: 17 (1 contradiction, 16 doubtful), each decided from the Arabic and fixed (scripts/recitals/ikhwan_fixes.py, logged in app/src/content/ai/ikhwan.json → fixes): ¶1 the sequence of fear and settlement; ¶2 the ḥaras/ḥarth note; ¶3 Bīrāst [Bēvarasp], Ṣāghūn [Balāsāghūn]; ¶9 «them [the animals]»; ¶17 «our» read as «[your]», with the print's reading stated; ¶19 «the intelligent»; ¶27 the Jews «hate us» with the print's reading stated, and the Rūm and Armenians as the topic of the next clause (the contradiction); ¶35, ¶45, ¶54, ¶57, ¶60, ¶68 («originated» for ikhturiʿat), ¶71. A person should check these.
11. Rights and private layers
C: PD in the US (1928), shown through H; H: CC BY-NC-SA 4.0 with attribution, as for the other OpenITI texts. B, Dieterici, Platts, Dowson: public domain. The IIS/OUP volumes, Goodman's 1978 translation, the French of de Vaulx d'Arcy and the Beirut edition: in copyright, cited only; nothing of their wording is in this reader and none is held privately. Audio: none made (about 9,200 Arabic and 11,000 English words: estimated about $2 on Azure; the author approves any TTS run above $1).
12. Open questions / to review
- A person should read the transcription's doubtful places (§4), above all ¶52 «الرتبة الكلية», in a better scan of C vol. 3 (HathiTrust was blocked by a bot check; worth a manual look) or in the IIS edition.
- Collate Part I with Dieterici's vocalised Arabic (1881); check which volume of BnF Arabe 6647–6648 holds Epistles 22 and 32.
- Acquire the IIS/OUP translations (Goodman & McGregor; Walker et al.; Baffioni) through a library into the private layer and rerun
triangulate.py ikhwan. - Epistles 1 and 5 (queued); the long middle of Epistle 22 (the delegations), if wanted.
13. Resonances (PROTOCOLS §16)
scripts/resonances.py whole --only ikhwan and passages ikhwan (batches of 4), with the author's pointers given to the model as an editor's hint (FOCUS in the script: consider, never assert); 19 new external references verified by an agent with web search, none removed (partial fits noted in the ledger: Goodman & McGregor 2009 for the Hebrew and Catalan afterlives, Hein 1985 for al-Kindī's Ḥudūd). Whole text 19 connections, passages 358 (253 parallels, 63 contrasts, 30 lineages, 12 shared sources). Most linked: al-Sijistānī's Maqālīd (47; the one of number and the One beyond it, the Command, Intellect and Soul), Ibn Sīnā's ʿIshq (35; tashabbuh, each being moved toward its perfection), Ibn Ṭufayl's Ḥayy, the Hayākil, Plotinus VI.9, the Symposium and the Republic. Graded: the definition of philosophy is a lineage (Theaetetus 176b through the Alexandrian prolegomena, Hein 1985); Plotinus's procession is a lineage through the Arabic Plotinus (Netton 1982; Adamson 2002); the Republic, the Symposium, the Timaeus, the Book of Causes and Nāṣir-i Khusraw are parallels (the last four whole-text ones curated, scripts/recitals/ikhwan_resonances_curated.py, because the generator cannot link the Rawshanāʾī-nāma, which has no English layer it reads); Plutarch's Gryllus, Montaigne, Bentham and the Zhuangzi are parallels to the animals' case, Descartes a contrast; Turmeda's Disputa de l'ase and Kalonymus's Hebrew version are its documented afterlives. On the Platonic Current the Ikhwān node now reads this text, with a faint thread to Plato (the Theaetetus definition, Hein 1985) and a convergence thread to the Book of Causes.
14. Kind, and one text in two parts
Kind: Philosophical treatises (philosophy), with the Niche of Lights, the Temples of Light and the Treatise on Love: argued philosophical prose on the order of being. Not "Ismaili philosophy" (al-Sijistānī, Nāṣir-i Khusraw), because the Ismaili authorship is contested (§0) and the kind would assert it; the Ismaili texts are linked from the reader and the resonances instead. Not "Visionary recitals", though Part I is a tale: the Brethren close it by saying it is no children's play but philosophy clothed in a story. One text, two parts (not two catalogue entries): one anonymous work that cites itself («as we have shown in our epistles»), one edition, one authorship question, one About page, one set of companions; the parts are its sections, as al-Sijistānī's fifteen keys are one text (the Maqālīd). The Fuṣūṣ chapters are separate entries because each is a whole chapter with its own commentaries and its own route; here each part opens as its own light page instead (§7), which keeps a phone's load to one epistle. A decision for the author: whether to list it under Ismaili philosophy as well (the catalogue has one kind per text).
15. Change log
- 2026-10-04: selection built from H (four epistles), cut by the author to two parts; collated against C (and B); 7 corrections; vowelled, analysed, translated; Dieterici, Platts and Dowson cut to the passages; triangulation; companions (64 items); version record; resonances; registered at
/read/ikhwan(two parts, one page each) under Philosophical treatises.