Patterns
The Qur'an's linguistic structure includes documented rhetorical features. Three are surfaced here: grammatical shifts (iltifat), similar phrasings (mutashabihat), and cross-references.
Iltifat: grammatical shift
Iltifat names a deliberate change in person, number, or addressee mid-passage. A canonical example is al-Fatihah:
ٱلْحَمْدُ لِلَّهِ رَبِّ ٱلْعَٰلَمِينَ ... إِيَّاكَ نَعْبُدُ
The first verses praise Allah in the third person; verse 5 turns to direct address ("It is You we worship"). This shift, studied at length by Abdel Haleem (1992), is treated as rhetorically deliberate, not a textual irregularity. See the shift in context: read al-Fatihah whole →
● al-Suyuti, al-Itqan.
Mutashabihat: similar phrasings
Many phrases recur across surahs with small variations. Recognizing these is a classical Qur'anic discipline ('ilm al-mutashabih). Example: فَلَا خَوْفٌ عَلَيْهِمْ وَلَا هُمْ يَحْزَنُونَ ("no fear shall be upon them, nor shall they grieve") appears in multiple surahs with subtle context differences.
Related-verse data is surfaced inline on the Read page at Study depth and above.
● phrase. Dataset: QUL mutashabihat (~5,277 entries), Tarteel.
Cross-references (Mishkat Mutashābihāt)
The cross-reference data shipped on this site — the expandable "Related verses" panels on the Read page and the browser below — is the Mishkat Mutashābihāt corpus: documented verse-to-verse parallels, browsable without tafsir traversal. Coverage: 110 of 114 surahs (not yet covered: 104, 105, 106, 110).
● dataset. See Sources.
Conceptual lineage: the QurSim corpus (Sharaf and Atwell, LREC 2012) pioneered this approach, identifying 7,679 pairs of related verses ranked by relatedness. QurSim is cited here as a conceptual reference only — its canonical host is no longer accessible, and it is not the data this site ships.
Browse cross-references by surah
The same Mishkat data, browsable a surah at a time: every verse in it that carries a documented parallel elsewhere in the Qur'an.
Sound and rhyme — the verse endings (fawāṣil)
Most surahs end their verses on recurring sound patterns — rhyme or assonance on the verse-final word (the fāṣila). Rhyme is one of the features scholarship reads structure from (Neuwirth's study of the Meccan surahs builds on it extensively). Pick a surah to see every verse ending, the rhyme families it groups into under two mechanical keys, where the ending pattern shifts, and any refrains (identical endings recurring three or more times).
~
Computed from the bundled morphology
(scripts/build-rhyme-map.mjs): each verse-final
word is reduced to an approximate pausal form (vowel marks
stripped — which also drops the case endings, as pause does —
final tā marbūṭa read as -ah), then keyed
two ways: fine (last two letters) and
coarse (final letter only). These are
orthographic proxies, not a phonological transcription of
recitation — endings that rhyme in recitation can split under
the fine key (the coarse key exists to show exactly that
tension). A shift in the key is a mechanical fact, not an
asserted structural boundary.
Rhyme regularity across surahs
The panel above shows one surah at a time; this ranks all 114 by mean run length — the average number of consecutive verses that share a fine rhyme key before it changes (verse count ÷ (shift count + 1)). A high value means long, sustained runs on one ending (near-monorhyme); a low value means the ending pattern changes almost every verse.
~ Leeds corpus v0.4.
ℹ How this is computed
~
computed purely from the fine-key shift counts already shown
above (data/rhyme-summary.json) — same
orthographic-proxy caveats apply, and a short surah reaching a
high value from very few verses is noted, not hidden.
Other documented features
- Fasilah: rhyming verse-end patterns, especially marked in early Meccan surahs.
-
Ring composition: chiastic structures where
openings and closings mirror. We tested one narrow, mechanical
proxy for this — whether a surah's rarest content roots
(occurring exactly twice) sit at matching distances from the
surah's start and end more often than random placement would
predict — across all 114 surahs (3,067 candidate root pairs,
an exact per-surah-length null distribution, Benjamini-Hochberg
correction at q<0.05). Zero pairs reached significance; the
single closest candidate still fell short of the corrected
threshold. This does not evaluate the literary scholarship
cited below, which analyzes thematic and semantic parallels,
not raw word position — it only shows this specific narrow
proxy carries no corpus-level signal.
Method and closest candidates
-
Block-level mirror symmetry: a second,
separate test asks the question the proxy above does not:
whether whole sections of a surah (mechanically
divided by where its vocabulary shifts most, not a scholar's
reading) mirror each other in content. Four tests per surah
(mirrored-section vocabulary overlap, first-section/
last-section overlap, recurring phrases bracketing both ends,
verse-length palindrome), corrected jointly across every
candidate in the corpus. Zero of 345 candidates reached
significance. Like the proxy above, a survivor here would
describe a mechanical property, never a literary finding,
and is never attributed to a scholar.
Method and closest candidates
- Parallelism: balanced clause structures.
- Qasam: oath openings (e.g., al-'Asr, al-Shams, al-Tin).
-
The believers' vocative:
ya ayyuha al-ladhina amanu ("O you who believe") opens
89 verses. This counts one specific formula — direct address in
the Qur'an (khitab) is far broader, reaching even
single pronouns ("It is You we worship"), and no
single count could capture it.
List all 89
-
Boundary particles and lexical continuity:
verses opening with idh ("when" / "recall"),
idha ("when" / "if" — same clause-initial temporal
family as idh, deliberately not extended to a
sequencing word like thumma, "then", which marks
continuation rather than a new temporal clause), or
lamma ("when" / "once") — 350 occurrences combined. In
137 of those, the verse shares a content root with the verse
immediately before it (e.g., 2:29 and 2:30 both use the root
a-r-ḍ, "earth"). This records only where the marker
and the shared root co-occur — it does not assert that the
boundary is a thematic or rhetorical turning point; that
judgment is an interpretive one this site does not make.
List all 137
● Abdel Haleem 1999.
ℹ How this is computed
● features documented in Abdel Haleem (1999), Neuwirth (1981), Cuypers (2015). ● The 89-verse count is computed directly from the bundled morphology (a match on the three consecutive lemmas of this formula), not asserted — see it listed above. ● The boundary-particle and shared-root counts are likewise computed directly from the bundled morphology (verse-initial lemma match, then a root intersection with the preceding verse) — see it listed above. ● The ring-composition proxy test (0 of 3,067 candidates significant) is likewise computed directly from the bundled morphology — method and closest candidates listed above.