Bazargan’s Chronological Method
How Bazargan arrived at an arrangement of 194 Blocks in 22 Groups. This account follows his working steps from the basic assumption through to the chronological groups, with the Persian sources quoted throughout.
1. The basic assumption: from empirical clue to chronological hypothesis
Bazargan’s method begins from a problem that is easy to state and hard to solve. The chronologies inherited from tradition and those compiled by modern scholars do not always give the same order. Bazargan therefore looked for a textual feature that could be counted directly and compared across surahs, without having first to adjudicate between conflicting reports.
Several candidates were available: rhyme, vocabulary, density of expression, the reports of asbāb al-nuzūl, and the relation of verses to historical events. Of all of these he settled on verse length, for an entirely practical reason — it is the feature most easily turned into numbers.
He writes:
«سادهتر و سرراستتر از همه اینها … بلندی و کوتاهی آیات است. یعنی تعداد کلماتی که هر آیه را تشکیل میدهد و ما آن را “طول آیه” مینامیم.»
“The simplest and most direct of all is the length of the verses, that is, the number of words that make up each verse, which we call ‘verse length’.” —Sayr-e Taḥavvol-e Qurʾān, I, p. 39.
The term “verse length” carries a narrow technical sense in Bazargan’s method: the number of words in a single verse unit according to his own counting system. The visual length of a line on the page of a muṣḥaf, the number of letters, and the number of syllables all fall outside the calculation.
1.1. Three surahs as a point of departure
Bazargan did not begin by counting the whole of the Qur’an and then announcing that verses grow longer over time. He took a more cautious route, first selecting three surahs whose relative chronological position he considered sufficiently agreed upon:
- Surah al-Humazah as an early Meccan surah;
- Surah Ṭā Hā as a surah from the middle Meccan period;
- Surah al-Māʾidah as one of the late Medinan surahs.
His initial figures were these:
| Surah | Period estimated by Bazargan | Words | Verses | Mean length |
|---|---|---|---|---|
| al-Humazah | First to second year of the mission | 34 | 9 | 3.78 |
| Ṭā Hā | Fourth to fifth year of the mission | 1,308 | 143 | 9.15 |
| al-Māʾidah | Late Medinan period | 2,749 | 120 | 22.91 |
The rise is unmistakable: 3.78 → 9.15 → 22.91.
From this came Bazargan’s first statement:
«با پیشرفت زمان، طول متوسط آیات سوره ترقی کرده است.»
“As time advances, the mean verse length of a surah increases.” —Sayr, I, p. 40.
He then compared several other surahs that are historically placed between Ṭā Hā and al-Māʾidah, such as al-Aʿrāf, al-Anfāl, al-Ḥashr, and al-Nūr. Their mean lengths also fell between those two points. The result gave him a clue that mean verse length might be related to the time of revelation.
1.2. A hypothesis still awaiting proof
At this stage Bazargan did not yet treat the relation as a settled law. His wording remains tentative:
«شاید بتوانیم حدس بزنیم که این قاعده کلیت داشته…»
“Perhaps we may surmise that this rule holds generally…” —Sayr, I, p. 41.
What he formulated instead was a question to be tested:
«آیا این تغییر نسبی تعداد کلمات آیات و تطویل اجمالی آنها، بر روال منظم و ثابت است و میتواند به عنوان قاعدهای اتخاذ شود یا خیر؟»
“Does this relative change in the number of words in the verses, and their general lengthening, proceed according to a regular and constant pattern such that it can be adopted as a rule?” —Sayr, I, p. 41.
The statement “verses grow longer over time” must therefore be understood as a working hypothesis. It was first obtained from a handful of surahs whose relative dating was taken as known, and then tested against the remaining surahs.
1.3. Why the mean alone is not enough
Bazargan recognised that two surahs may have almost the same mean length while being composed of quite different verse distributions. One surah may contain many verses of closely similar length, while another mixes very short verses with very long ones. Both can yield a similar mean.
He therefore did not stop at mean length. He drew up a distribution table showing how many verses had a length of 3 words, 4 words, 5 words, and so on.
The comparison of his three initial surahs produced the following pattern:
| Characteristic | al-Humazah | Ṭā Hā | al-Māʾidah |
|---|---|---|---|
| Mean length | 3.78 | 9.15 | 22.91 |
| Dominant length | 4 words | 5 words | 13 or 15 words |
| Verses at the dominant length | 55% | 13.3% | 7.5% |
| Number of distinct lengths | 3 | 22 | 39 |
| Range of verse lengths | 3–5 words | 2–27 words | 7–64 words |
| Shape of the curve | Tall and narrow | Intermediate | Low and wide |
The table shows that the chronological change Bazargan had in mind involves three movements occurring together, of which the rise in the mean is only one.
First, the centre of the distribution shifts to the right. The dominant length rises from 4 words in al-Humazah to 5 words in Ṭā Hā, and then to 13 or 15 words in al-Māʾidah.
Second, the concentration at the dominant length declines. In al-Humazah, 55 per cent of the verses share the dominant length. In Ṭā Hā the figure falls to 13.3 per cent, and in al-Māʾidah to only 7.5 per cent. Since the vertical axis of the curve records the percentage of verses, this loosening of concentration reads directly as a progressively lower peak.
Third, the range of the distribution grows wider. Al-Humazah has only three distinct verse lengths — 3, 4, and 5 words. Ṭā Hā has 22, and al-Māʾidah 39. This is what is meant by saying that the curves of the later surahs “widen”.
Bazargan summarises his findings:
«طول غالب زیاد شده است، درصد طول غالب تنزل کرده است، و دامنه آیات گسترش یافته است.»
“The dominant length has increased, the percentage at the dominant length has fallen, and the range of the verses has widened.” —Sayr, I, p. 55.
In his graphical description he writes:
«منحنی همزه نوکتیز و مرتفع است و منحنی مائده خفته و گسترده است.»
“The curve of al-Humazah is sharp-peaked and tall, while the curve of al-Māʾidah is low and spread out.” —Sayr, I, p. 57.
It should be stressed that “more uniform” here does not mean that all early verses have the same number of words. What it means is that a greater proportion of their verses is concentrated in a few length categories. The later surahs display greater variation in verse length, so that their distribution is more dispersed.
1.4. Bazargan’s hypothesis stated more precisely
Bazargan’s statistical hypothesis can be set out as four tendencies:
- The mean verse length tends to increase over time.
- The dominant length, or mode, tends to shift towards a larger number of words.
- The percentage of verses at the dominant length tends to fall, so that the peak of the curve becomes lower.
- The range of variation in verse length tends to widen, so that the curve becomes broader and extends further to the right.
The chronological marker is therefore a joint change across several characteristics of the distribution at once, and no single figure will capture it. In the early stage the curve is tall, narrow, and sits to the left; in the later stage it is lower, wider, and displaced to the right.
It is this joint movement that Bazargan later sought to condense through mean length, dominant length, curve height, and the range of the distribution, before combining them into the ṭūl-e mabnā index.
1.5. A general tendency with local fluctuation
Although he uses the term “rule of verse lengthening”, Bazargan does not claim that every surah revealed later must have longer verses than the surah immediately preceding it. He distinguishes the long-run tendency from individual fluctuation.
He writes that it presents no difficulty if:
«سیر تحول طول متوسط آیات در جهت ترقی کلی در طی زمان بوده باشد، ولی از یک سوره به سوره بعدی دچار کم و زیادی گردد…»
“The development of the mean verse length moves in a generally rising direction over time, while from one surah to the next it may rise and fall…” —Sayr, I, p. 65.
He even concedes that verse length:
«نتواند معرف تقدم و تأخر دقیق حتمی سورهها باشد.»
“cannot serve as a certain indicator of the exact precedence and succession of the surahs.” —Sayr, I, p. 65.
Bazargan’s basic assumption must therefore not be formulated as “every later revelation always has longer verses”. The more accurate formulation runs:
Over the period of revelation there is a long-run statistical tendency towards longer verses and a more varied distribution of lengths. That tendency can be used to estimate relative chronological position, though it does not fix the exact order of every surah or Block.
This distinction matters for MushafNuzuli. Bazargan’s model looks for the general direction of development within the corpus, and stops there. It promises no deterministic law holding without exception for every unit of revelation.
2. Tokenisation
The hypothesis above requires something that must be supplied first, namely numbers. Before verse length can be counted, the text of the Qur’an has to be turned into countable units, and this process is called tokenisation.
Tokenisation determines which part of the Arabic text counts as one statistical word. The process is needed because word boundaries in Arabic cannot always be settled by spaces alone. Some particles are written attached to another word, while others are written separately even though grammatically they bind closely to the word that follows.
Bazargan’s first question is therefore more specific than simply “how many space-separated forms are there in this verse?”. What he asks is:
“According to the counting rules in use, how many word units are there in this verse?”
2.1. Why tokenisation rules are needed
Bazargan found that the muṣḥaf editions he used did not always apply the same method of counting words and verses. He explains:
«چون دیده شد که در شمارش آیات و کلمات، رویه ثابت مشترک خالی از اختلاف و اشتباه … وجود ندارد، رویه ثابتی اتخاذ گردید.»
“Since it was found that in the counting of verses and words there exists no common, fixed practice free of discrepancy and error, a consistent practice was adopted.” —Sayr-e Taḥavvol-e Qurʾān, I, p. 47, note 1.
Bazargan thus did not simply accept the word counts printed at the head of a surah. He established his own operational rules and then applied them consistently across the whole corpus.
The aim of those rules is to produce a uniform unit of counting so that verse lengths from different surahs can be compared. Producing a flawless morphological analysis of Arabic lies outside their purpose.
2.2. Bazargan’s tokenisation rules
His main rules can be summarised as follows:
| Linguistic form | How it is counted |
|---|---|
| A single conjunctive, negative, prohibitive, or interrogative particle | Joined to the word it relates to, without being counted separately |
| Two particles occurring in sequence | The string of particles and the following word counts as two tokens |
| Words such as huwa, idhā, and hādhā | Counted as independent tokens |
| Mā when not functioning as a negative particle | Counted as an independent token |
Bazargan lists several particles that are not counted separately:
«حروف عطف و ربط و نفی و نهی و استفهام … مانند: و، ف، لا، لم، هل»
“Particles of conjunction, connection, negation, prohibition, and interrogation, such as wa, fa, lā, lam, and hal.”
Such single particles are treated as part of the word they relate to. When two particles occur in sequence, Bazargan counts the string as two words. The examples he gives include:
- fa-lā taqul;
- a-lam taʿlam;
- a-fa-lā taʿqilūn;
- a-fa-ghayr;
- law-lā kāna;
- yā ayyuhā alladhīna.
These rules are Bazargan’s own statistical convention, and they are not always identical with the division of words according to Arabic morphology, muṣḥaf orthography, or modern digital corpus tokenisation.
2.3. An example of the tokenisation process
Consider the opening of Surah al-Ghāshiyah:
هَلْ أَتَاكَ حَدِيثُ الْغَاشِيَةِ
If the text is divided by spaces alone, four units result:
هَلْ · أَتَاكَ · حَدِيثُ · الْغَاشِيَةِ
Under Bazargan’s rules, hal is an interrogative particle and is not counted as an independent word. It is treated together with the word that follows:
هَلْ أَتَاكَ · حَدِيثُ · الْغَاشِيَةِ
The statistical length of that verse is therefore 3, and not four words.
The word huwa receives different treatment, since it counts as an independent token. In the verse:
قُلْ هُوَ اللَّهُ أَحَدٌ
the tokenisation becomes:
قُلْ · هُوَ · اللَّهُ · أَحَدٌ
so that the verse length is 4.
These examples show that a token in Bazargan’s method is a counting unit established by convention, and does not always coincide with a word visually separated by spaces.
2.4. The stages of tokenisation
In simple terms, Bazargan’s tokenisation can be described in four stages.
-
Reading the text of each verse. Every verse is treated as one basic unit whose length is to be measured.
-
Identifying the linguistic forms. Particles such as wa, fa, lā, lam, and hal are distinguished from words counted independently.
-
Applying the joining rules. Single particles are joined to the related word, while a sequence of two particles is treated under Bazargan’s two-token rule.
-
Recording the verse token count. Once the whole text of a verse has been processed, its token count is recorded as the length of that verse.
The immediate result of tokenisation is not yet a date or a chronological position. What is obtained is a series of verse-length figures — 4, 5, 4, 7, 6, 5, and so on.
It is this series that is then used to compute:
- the total word count of the unit;
- the mean verse length;
- the most frequently occurring length;
- the distribution of verse lengths;
- and the height and shape of the characteristic curve.
2.5. Why consistency matters more than linguistic perfection
Bazargan was aware that his tokenisation rules are open to dispute from the standpoint of Arabic grammar. In his view the problem does not damage the comparison as long as the same rules are applied to every surah:
«چون این طرز عمل در همه جا به طور یکسان اجرا شده است و محاسبات و مطالعات بهصورت نسبی و مقایسه است … ایجاد تبعیض و خطا نمیکند.»
“Since this procedure has been carried out uniformly everywhere, while the calculations and studies are relative and comparative in character, it introduces no distortion into the comparison.” —Sayr, I, p. 47, note 1.
His argument is that this chronological method requires above all a stable system of counting, and that a linguistically incontestable definition of the word ranks second in importance. If the whole text is processed under the same rules, the results for different surahs and Blocks remain comparable within the model.
Consistency nevertheless does not remove every problem. Differences in tokenisation can alter:
- the word count of a verse;
- the mean length;
- the dominant length;
- the height of the curve;
- the value of ṭūl-e mabnā;
- and ultimately the relative position of a Block.
Bazargan’s word counts must therefore not be mixed directly with counts from Tanzil, the Quranic Arabic Corpus, or space-based counting without first establishing that the tokenisation rules are the same.
2.6. Tokenisation and segmentation work at different levels
| Process | Question answered | Result |
|---|---|---|
| Tokenisation | What counts as one word? | The token count of each verse |
| Segmentation | Which verses form one textual section? | Thematic or structural Segments |
| Block formation | Which sections are taken to come from the same phase of revelation? | Chronological Blocks |
Tokenisation applies to the text of the verse. Segmentation then divides the surah according to changes in theme, style, rhyme, or statistical character. Several Segments may subsequently be combined into a single statistical unit or a single chronological Block.
The method as a whole therefore runs as follows:
text of the Qur’an → tokenisation → length of each verse → initial distribution → segmentation and Block formation → Block distribution → ṭūl-e mabnā
Tokenisation supplies the data to be counted. Segmentation determines the boundaries of the units whose data will be compared — and that is the subject of the next section.
3. Segmentation
Tokenisation determines what counts as one word and yields the token count of each verse. Bazargan’s method still does not always treat an entire surah as a single chronological unit. Before the final distribution and the value of ṭūl-e mabnā are computed, he must decide which verses deserve to be analysed together.
The process of establishing the boundaries of these sections of a surah is called segmentation. The question it answers is:
Do all the verses in this surah form a single, relatively homogeneous statistical unit, or does the surah consist of several parts that need to be computed separately?
The following four levels of work help to keep processes apart that are easily confused:
| Process | Object | Question | Result |
|---|---|---|---|
| Tokenisation | The text of each verse | What counts as one word? | The length of each verse |
| Segmentation | The arrangement of verses within a surah | Which verses form one textual section? | Segments |
| Statistical unit formation | The Segments already identified | Which Segments deserve to be counted together? | Statistical units |
| Block formation | The units being compared | Which sections are taken to come from the same phase? | Chronological Blocks |
3.1. Why a surah is not always counted as a single whole
Mean length and the distribution curve are meaningful only when the verses counted together come from a relatively comparable population. If the opening of a surah is dominated by short verses while its closing part is dominated by long ones, the mean for the whole surah becomes a composite figure representing neither.
Bazargan explains:
«اگر طرز توزیع یا “پاشش آیات” بهلحاظ طولهای مربوطه در قسمتهای مختلف سوره یکسان نباشد، طول متوسطی که حساب میشود و شاخصهایی که از منحنی درمیآیند، جنبه مصنوعی و قراردادی پیدا کرده و معنای صحیح خود را از دست میدهد.»
“If the pattern of distribution, or the ‘scatter of the verses’ by their respective lengths, is not the same across the different parts of a surah, then the mean length that is calculated and the indicators derived from the curve take on an artificial and conventional character and lose their proper meaning.” —Sayr-e Taḥavvol-e Qurʾān, I, p. 72.
Imagine a hypothetical surah with two parts:
| Part | Verse lengths |
|---|---|
| First part | 4, 5, 5, 6, 5 |
| Second part | 15, 17, 18, 16, 19 |
The mean for the whole surah is (4+5+5+6+5+15+17+18+16+19) ÷ 10 = 11.
Hardly any verse is anywhere near 11 words long. The figure simply sits between two different distributions: the mean of the first part is 5, while the mean of the second is 17.
Counting the whole surah as one unit in a case like this would mask its internal difference. Bazargan therefore proposes:
«در چنین حالت، بهتر است تفکیک قائل شده، برای هر قسمت سوره که پاشش یکنواخت داشته باشد حساب جداگانه درنظر بگیریم.»
“In such a case it is better to make a division and to consider a separate calculation for each part of the surah that has a uniform scatter.” —Sayr, I, p. 72.
3.2. What Bazargan means by taqṭīʿ
Bazargan calls this procedure taqṭīʿ-e sūrah-hā, the “cutting” or segmentation of surahs. His operational account of it appears in a dedicated section headed «تقطیع سورهها»:
«سوره را برحسب آیاتی که درباره موضوع واحد و مرتبط با یکدیگر هستند، دستهبندی میکنیم. سپس برای هر دسته، طول متوسط و طول غالب را تعیین کرده، دسته آیات را بهلحاظ طولهای متوسط و غالب با هم میسنجیم.»
“We group the surah according to those verses that treat a single subject and are connected with one another. Then, for each group, we determine the mean length and the dominant length, and we compare the groups of verses in terms of their mean and dominant lengths.” —Sayr, I, p. 87.
This passage shows that Bazargan’s segmentation proceeds in two stages.
First, the surah is divided according to textual and thematic connection. Verses forming a single discourse, narrative, argument, command, or rhetorical sequence are placed in one candidate Segment.
Second, each Segment is tested statistically. Its token count, verse count, mean length, and dominant length are computed. The Segments are then compared to see whether a thematic difference is accompanied by a difference in verse-length character.
Bazargan thus does not run a statistical algorithm that finds cut-points automatically. Theme is used to propose Segment boundaries, while statistics are used to check whether those boundaries produce a meaningful difference.
3.3. The clues used to set Segment boundaries
In practice Bazargan draws on several kinds of clue at once.
A change of theme or discourse
A shift from a prophetic narrative to eschatological warning, from theological polemic to legal provision, or from narrative to prayer may all be grounds for opening a new Segment.
A change of theme does not by itself establish a difference in the time of revelation. It merely supplies a provisional boundary to be tested.
The connection between verses
Verses forming a single argument or narrative sequence should not be separated merely because their means differ slightly. Bazargan several times rejects divisions he regards as damaging to continuity of meaning.
In his discussion of al-Muzzammil, for instance, he rejects one of Blachère’s divisions because the verses still form a sequence of mutually completing commands and share the same rhyme:
«بنابراین یا سوره مزمل در ۱۹ آیه اول خود یکپارچه است، کما اینکه یک قافیه است…»
“Accordingly the first nineteen verses of Surah al-Muzzammil form a single whole, just as they also share a single rhyme.” —Sayr, I, p. 77.
Reports and historical information
A report stating that a verse was revealed in connection with a particular event may be grounds for examining it separately. Bazargan cites the verse of mubāhalah, the verse of taṭhīr, and the verse of tablīgh as examples of verses traditionally linked to a specific context.
Such reports are nevertheless not accepted without testing. He weighs them against the context of the verse and its statistical character.
A change in the distribution of verse lengths
If a particular part of a surah is consistently concentrated at a length different from the rest, that is a sign that the surah as a whole may not be homogeneous.
Bazargan describes the ideal case as a relatively similar distribution across the beginning, middle, and end of a surah. If the opening rests on the short-verse categories while the closing part moves to the long ones, a separate calculation comes into consideration.
Outlying verses
A single verse far longer or shorter than everything around it can alter the mean and the shape of the curve. Such a verse is examined to see whether it is an integral part of the surah or an addition from another phase.
The clearest example is the final verse of al-Muzzammil. Its length differs sharply from the preceding nineteen verses, and its content presupposes a more developed community. Bazargan therefore counts it as a unit of its own.
Rhyme and rhetorical form
A change of rhyme can support a boundary, while continuity of rhyme can support unity. Rhyme nevertheless remains no sole criterion, and is used together with theme, context, and the results of calculation.
3.4. From Segment to statistical unit
Once candidate Segments have been formed, the mean length of each is computed in a straightforward way: the total token count of the verses in the Segment divided by the number of verses in it.
The dominant length or mode of each Segment is recorded as well. Segments with a textual connection and a similar statistical character may be combined into a single statistical unit. Conversely, two textually adjacent Segments are not always combined if the difference in their statistics and content is judged strong enough.
In MushafNuzuli’s terminology:
- a Segment is the initial division of a surah according to the structure of the discourse;
- a statistical unit is one or more Segments whose data are computed together;
- a Block is the final unit taken to represent one phase of revelation.
Bazargan himself does not always use the corresponding Persian terms with the rigour of a database schema. This three-level distinction is an editorial standardisation by MushafNuzuli, adopted so that the process can be described consistently.
3.5. Segmentation does not always end in dividing the surah
Forming Segments is still a stage of testing. After comparing several Segments, Bazargan may conclude that the differences are not strong enough and restore the surah to a single whole.
An example is Surah al-Naml, which he divides into seven thematic Segments:
| Segment | Verses in Bazargan’s numbering | Theme | Mean | Dominant |
|---|---|---|---|---|
| 1 | 1–6 | The Qur’an as guidance | 7.17 | 9 |
| 2 | 7–14 | The mission of Moses | 12.88 | 11 |
| 3 | 15–45 | Sulaymān and the Queen of Sabaʾ | 13.13 | 11 |
| 4 | 46–54 | Thamūd and Ṣāliḥ | 10.11 | 9 |
| 5 | 55–59 | The story of Lūṭ | 9.40 | 8 |
| 6 | 60–77 | Divine unity, resurrection, and prophethood | 12.71 | 9 |
| 7 | 78–95 | The Qur’an, resurrection, and prophethood | 12.00 | 11 |
This numbering follows the mixed system Bazargan adopted and is therefore not identical with the Kufan numbering used in most muṣḥaf editions today.
The first and fifth Segments appear shorter than the rest. Each consists of only six and five verses respectively, and Bazargan judged so small a sample too unstable to support a chronological division. He also considered it unlikely that the stories of Lūṭ and Ṣāliḥ were revealed separately from the sequence of prophetic narratives around them.
His conclusion:
«بنابراین سوره را یکپارچه و با آیات معاصر میگیریم.»
“We therefore take the surah as a single whole, with its verses contemporaneous.” —Sayr, I, p. 89.
The example matters because it shows that a large number of thematic Segments does not by itself produce a large number of chronological Blocks. A surah may have several themes and still be taken to come from a single phase of revelation.
3.6. When segmentation does yield several Blocks
A different outcome appears in the analysis of Surah Ibrāhīm. Bazargan first divides it into eight thematic Segments. Several adjacent Segments are then combined, producing five statistical units.
He writes:
«هشت دسته فوق که تشکیل ۵ مجموعه جدید میدهند، تماماً دارای طولهای غالب و متوسط نزدیک مشابه بوده، ممکن است معاصر گرفته شوند؛ به استثنای دسته ۸ یا واحد پنجم…»
“The eight groups above, which form five new units, all have dominant and mean lengths that are closely similar and may be taken as contemporaneous — with the exception of the eighth group, or fifth unit.” —Sayr, appendix, p. 275.
In MushafNuzuli’s presentation the structure can be summarised as: 8 thematic Segments → 5 statistical units → 3 chronological Blocks.
Segments 1–7 generally have closely similar mean lengths and are placed in Block A. The eighth Segment has a lower mean and dominant length and is made into Block B. Certain verses regarded as later additions are then placed in Block C.
The case of Surah Ibrāhīm shows that Segments, statistical units, and Blocks are not interchangeable:
- eight Segments do not produce eight Blocks;
- several Segments may be combined before the final curve is drawn;
- one Block may encompass a number of Segments;
- and non-adjacent verses may also be placed in a single Block if they are taken to be contemporaneous.
The details of the verse boundaries and their computation call for separate treatment in a chapter devoted to Surah Ibrāhīm, including the arithmetical inconsistencies in some of Bazargan’s tabulated figures.
3.7. The segmentation procedure as a whole
Bazargan’s procedure can be reconstructed as follows.
-
Compute the length of each verse. Tokenisation yields the series of verse lengths.
-
Test the surah as a single whole. The mean length, mode, distribution, and curve of the entire surah are computed first.
-
Look for internal non-uniformity. Bazargan checks whether the beginning, middle, and end of the surah have different length scatters or contain outlying verses.
-
Form candidate Segments. Verses are grouped by theme, narrative connection, rhyme, historical context, and possible textual boundaries.
-
Compute the statistics of each Segment. Word count, verse count, mean, and dominant length are computed for each.
-
Compare the Segments. Segments with similar statistics may be combined into statistical units. Differences arising from only two or three verses are treated with caution.
-
Test the result of combination. The distribution and curve of the combined unit are recomputed. Bazargan compares whether the result is more regular and its indicators better aligned than those of the whole surah.
-
Establish the Blocks. If the textual and statistical differences are judged strong enough, the units are established as distinct groups of contemporaneous verses.
The full sequence becomes:
tokenisation → verse lengths → test of the whole surah → candidate Segments → statistical units → Blocks → Block distribution curve → ṭūl-e mabnā
3.8. Segmentation as an interpretive decision
Segmentation is one of the most interpretive parts of Bazargan’s method. Its boundaries are not produced by any single formula. Bazargan moves repeatedly between reading the text and reading the statistics: a thematic boundary is proposed, the calculation is run, the curve is assessed, and the boundary is revised if need be.
The procedure has the strength of not relying on figures in disregard of content. It also carries the risk of circular reasoning. An irregular curve becomes the reason for cutting a surah, and the regularity of the curve after the cut is then used to justify the cut.
Several further limitations deserve note:
- small Segments yield unstable means and modes;
- several possible segmentations may produce curves that all look equally plausible;
- the judgement that a theme is “different” is not wholly objective;
- the choice of verse-numbering system can alter the boundaries and the number of statistical units;
- and a more regular curve does not by itself prove a difference in the time of revelation.
MushafNuzuli should therefore present Segment and Block boundaries as the result of Bazargan’s reconstruction, and refrain from displaying them as a demonstrated original structure of the Qur’an.
The most accurate formulation is:
Segmentation is the process of dividing a surah into candidate textual sections, then testing whether those sections have verse-length characteristics similar or different enough to be counted as one or several chronological units.
Segmentation supplies the units to be analysed, and stops there. Their chronological position is obtained only once the Block distributions have been computed, the curves drawn, and the value of ṭūl-e mabnā determined.
4. The distribution of verse lengths
Tokenisation yields the token count of each verse, and segmentation fixes the boundaries of the Blocks to be analysed. The next stage is to draw up the distribution of verse lengths, which answers the question:
Within a given Block, how many verses consist of 3 tokens, 4 tokens, 5 tokens, and so on?
Attention at this stage moves from the position of each verse to the frequency with which each length category occurs. Bazargan explains:
«اگر آیات یک سوره را تکتک، کلمهشماری و جدولی از آنها تنظیم نماییم، چهره تازهای از هر سوره ظاهر میشود.»
“If the verses of a surah are counted word by word and arranged into a table, a new face of each surah appears.” —Sayr-e Taḥavvol-e Qurʾān, I, p. 52.
Bazargan first applied this procedure to whole surahs. Once composite surahs had been divided, the same table was applied to groups of verses regarded as homogeneous. In MushafNuzuli’s terminology, that unit of distribution is the Block.
4.1. From a series of verse lengths to a distribution
Suppose a Block consists of N verses. Tokenisation gives the series of lengths l1, l2, l3, through to lN, where li is the token count of verse i.
As an example, a hypothetical Block yields the series 3, 4, 4, 5, 3, 4, 6, 4.
That series still preserves the order of the verses. To form a distribution, the verses are grouped by their length:
| Length k | Verses of that length | Frequency fk |
|---|---|---|
| 3 tokens | Verses 1 and 5 | 2 |
| 4 tokens | Verses 2, 3, 6, and 8 | 4 |
| 5 tokens | Verse 4 | 1 |
| 6 tokens | Verse 7 | 1 |
Formally, fk is the number of verses having exactly k tokens. Since every verse must fall into one of the length categories, the sum of all fk equals N.
The number of tokens contained in one category is k multiplied by fk, while the total token count of the whole Block — written W — is the sum of all such products.
A distribution is therefore a change in the mode of presentation: from “verse → length of each verse” to “a given length → number of verses of that length”.
4.2. Absolute and relative frequency
Bazargan records two kinds of measure for each length category. The first is the absolute frequency, that is, the verse count fk. The second is the relative frequency pk, the percentage of verses at that length, obtained by dividing fk by N and multiplying by one hundred per cent.
In the hypothetical example above, four of the eight verses have a length of 4 tokens, so p4 is 50 per cent.
Using percentages makes it possible to compare Blocks with different verse counts. A frequency of 10 verses looks large in a Block of 20 verses and small in a Block of 100 — 50 per cent in the first case, 10 per cent in the second. The vertical axis of Bazargan’s curve therefore uses the percentage of verses, and not the raw verse count.
4.3. The percentage of verses and the percentage of tokens are not the same
Bazargan’s tables also record the percentage of all tokens contained in each length category, written qk, obtained by multiplying k by fk, dividing by W, and multiplying by one hundred per cent.
This measure answers a different question:
| Measure | Question |
|---|---|
| pk | What percentage of verses has length k? |
| qk | What percentage of all tokens lies in verses of length k? |
Because qk multiplies the frequency by the verse length, the categories of long verses receive greater weight. This measure must therefore not be used in place of pk when reconstructing Bazargan’s curve. The characteristic curve is built from the pairs (k, pk).
4.4. An example: the distribution of Surah al-Humazah
The simplest example Bazargan gives is Surah al-Humazah. Under his tokenisation rules, the nine verses of that surah are grouped as follows:
| Length | Verse numbers | Verses | Tokens | Percentage of verses | Percentage of tokens |
|---|---|---|---|---|---|
| 3 tokens | 6, 8, 9 | 3 | 9 | 33.3% | 26.5% |
| 4 tokens | 1, 2, 3, 5, 7 | 5 | 20 | 55.6% | 58.8% |
| 5 tokens | 4 | 1 | 5 | 11.1% | 14.7% |
| Total | 9 | 34 | 100% | 100% |
The mean length is 34 divided by 9, that is 3.78. The category with the highest frequency is 4 tokens with five verses, so the dominant length or mode is 4. The percentage of verses at that dominant length is 5 divided by 9, that is 55.6 per cent.
The example brings out the difference between the mean and the dominant length. The mean length of al-Humazah is 3.78 tokens, yet not a single verse is 3.78 tokens long. The mean is the result of dividing all the tokens, while the mode is the length category actually encountered most often.
4.5. The mean length as a product of the distribution
The mean length can be computed directly as W divided by N. The same figure can also be obtained from the distribution table, by summing all the products of k and fk and dividing by the sum of all fk. If percentages are what is available, the mean follows from summing all the products of k and pk divided by one hundred.
The distribution table is therefore not separate from the mean length. It shows precisely how that mean comes about.
Two Blocks may share the same mean with different distributions. The series 2, 2, 8, 8 and the series 4, 5, 5, 6 both yield a mean of 5. The first Block mixes very short with very long verses, while the second is concentrated around 5 tokens. That difference is visible in the distribution and hidden in the mean.
4.6. Forming the characteristic curve
Once the relative frequency of each length has been obtained, Bazargan transfers it to a coordinate plane. He explains:
«محور افقی را … برحسب تعداد کلمات درجهبندی میکنیم و محور قائم را … برحسب درصد آیاتی که دارای آن عده کلمه هستند، درجهبندی میکنیم.»
“We graduate the horizontal axis according to the number of words, and the vertical axis according to the percentage of verses having that number of words.” —Sayr, I, p. 56.
The horizontal axis thus shows verse length in tokens, while the vertical axis shows the percentage of verses at that length. For al-Humazah the points are (3, 33.3), (4, 55.6), and (5, 11.1).
Those points are then joined. Bazargan calls the resulting line the characteristic curve (monḥanī-ye moshakhkhaṣeh):
«اگر آن نقاط را بهیکدیگر وصل کنیم یک خط منکسر مضرس بهدست میآید که آن را منحنی مشخصه آن سوره میگوییم.»
“If we join those points to one another, we obtain a broken, serrated line, which we call the characteristic curve of that surah.” —Sayr, I, pp. 56–57.
Although it is called a “curve”, the data on the horizontal axis are in fact discrete. A verse 4.5 tokens long cannot exist. The connecting line is merely a visual convention that makes the overall shape of the distribution easier to see.
4.7. What can be read from the curve
The distribution curve displays several characteristics at once.
Horizontal position
The position of the curve shows the length category in which most verses are gathered. A curve lying further to the right is generally made up of longer verses.
The peak of the distribution
The length category with the largest percentage produces the peak of the curve, and the position of that peak indicates the dominant length. If two categories share the highest frequency, the distribution has two modes — Bazargan, for instance, gives the dominant length of al-Māʾidah as 13 or 15 words.
The height of the peak
The height of the peak shows how far the verses are concentrated at the dominant length. A tall peak means that many verses share the same or a closely similar length, while a low peak indicates that the verses are spread across more categories.
The width of the distribution
The width of the distribution shows the variation in verse length. Al-Humazah has verses of only 3, 4, and 5 tokens, whereas other surahs may range from very short verses to verses of several dozen tokens.
This observed range of lengths must be distinguished from the formal dāmaneh index that Bazargan introduces later. The dāmaneh index is a particular measure taking account of the widening, the asymmetry, and the height of the curve, and so amounts to more than the difference between the largest and smallest length. The two must not be interchanged.
Tails and asymmetry
If the curve extends far to the right, the Block contains a number of verses much longer than its dominant length. Those long verses form a right-hand tail and can raise the mean without moving the mode.
4.8. The original curve and the equivalent curve
A curve formed directly from the percentages of each length category is usually serrated. Some categories may have a high frequency, the next low or zero, and the one after that higher again.
To reduce this irregularity, Bazargan draws a second curve, which he calls the equivalent characteristic curve:
«برای آنکه از اثر تضاریس کاسته شود … بهجای منحنی مضرس حفرهدار، یک منحنی منظمی که معادل آن باشد رسم کنیم.»
“In order to reduce the effect of the ruggedness and obtain a simpler form, in place of the serrated and pitted curve we may draw a regular curve equivalent to it.” —Sayr, I, p. 59.
The equivalent curve is drawn through the middle of the fluctuations, so that the portions lying above and below the original curve balance one another. Bazargan requires the area beneath the two curves to remain equal.
Three things must therefore be kept distinct:
| Form | Basis of formation | Function |
|---|---|---|
| Original curve | The empirical points (k, pk) | Displays the distribution as actually observed |
| Equivalent curve | Smoothing of the original curve | Displays the overall tendency of the shape |
| Statistical indicators | Taken from the table or from a particular curve | Serve as inputs to the ṭūl-e mabnā calculation |
Bazargan’s smoothing is done graphically, without any explicitly formulated statistical algorithm. Two researchers may therefore draw slightly different equivalent curves. In MushafNuzuli’s digital reconstruction, the original curve must always be preserved as the primary data, while the equivalent curve needs to be labelled as the product of smoothing.
4.9. What is lost when a distribution is formed
A distribution preserves the frequency of lengths and discards the order of the verses. The series 3, 4, 4, 5, 6 and the series 6, 4, 3, 5, 4 produce the same distribution although the arrangement of their verses differs.
A distribution therefore cannot show:
- whether the long verses are gathered at the end of a Block;
- whether the lengths change gradually;
- the narrative connection between verses;
- changes of theme or rhyme;
- or whether a given verse is an outlier at a particular position.
This is why segmentation must be carried out with the order and structure of the text still in view. Once the distribution has been formed, that information about order is no longer visible in the curve.
4.10. The effect of Block size
The distribution of a small Block is less stable than that of a large one. If a Block has only nine verses, as al-Humazah does, a single verse represents 11.1 per cent. Reclassifying one verse alone can change the height of the curve by 11.1 percentage points. In a Block of 100 verses, one verse alters the distribution by a single percentage point.
In small Blocks, consequently:
- the peak shifts easily;
- the mode changes readily or becomes double;
- the curve appears heavily serrated;
- one long verse can strongly affect the mean;
- and the smoothing of the curve depends more heavily on the researcher’s judgement.
Small differences between two Blocks must therefore not be read straight off as evidence of a chronological difference, particularly where the verse counts are low.
4.11. The result of the distribution stage
The immediate result of this stage is one table per Block, recording the pairs of length, frequency, and percentage. No date of revelation appears at this point.
From that table are obtained:
- the verse count N;
- the token count W;
- the mean length;
- the dominant length L;
- the percentage at the dominant length;
- the shape of the distribution;
- the original characteristic curve;
- and, in Bazargan’s method, the equivalent characteristic curve.
The methodological sequence becomes:
tokenisation → length of each verse → segmentation → Blocks → Block distribution table → characteristic curve → statistical indicators → ṭūl-e mabnā
The distribution of verse lengths can therefore be formulated as:
The process of grouping the verses of a Block by their token counts, computing the frequency and percentage of each length category, and presenting the result as a table and a characteristic curve.
A distribution does not by itself determine chronology. It converts verse-length data into a form in which the statistical character of different Blocks can be compared. Chronological value is assigned only at the next stage, through the indicators of the curve and the calculation of ṭūl-e mabnā.
5. The characteristic curve
A distribution table shows how many verses fall into each length category. When many Blocks are to be compared, the rows of figures in a table are not always easy to read. Bazargan therefore converts the distribution into a geometrical form he calls the characteristic curve (monḥanī-ye moshakhkhaṣeh).
The characteristic curve answers the question:
How are the verses of a Block spread from the short categories to the long ones, and in which length category are they most concentrated?
The curve produces no new data. It gives visual form to data already present in the distribution table: distribution table → coordinate points → characteristic curve.
In MushafNuzuli’s terminology, every Block has its own distribution table and its own characteristic curve.
5.1. From distribution to coordinate points
Suppose a Block has N verses. For each length category k, one computes fk, the number of verses whose length is k tokens. The percentage of verses in that category, pk, follows from dividing fk by N and multiplying by one hundred per cent.
Each length category is then turned into a point with the coordinates (k, pk): the horizontal coordinate gives the number of tokens in the verse, and the vertical coordinate gives the percentage of verses of that length.
Bazargan explains:
«محور افقی را … برحسب تعداد کلمات درجهبندی میکنیم و محور قائم را … برحسب درصد آیاتی که دارای آن عده کلمه هستند، درجهبندی میکنیم.»
“We graduate the horizontal axis according to the number of words, and the vertical axis according to the percentage of verses having that number of words.” —Sayr-e Taḥavvol-e Qurʾān, I, p. 56.
If 30 per cent of the verses in a Block are 4 tokens long, the distribution yields the point (4, 30) — placed directly above the figure 4 on the horizontal axis and level with 30 per cent on the vertical.
5.2. Why percentages are used
The vertical axis uses the percentage of verses rather than the raw verse count. This is necessary because Blocks differ in the number of verses they contain. A category made up of 10 verses may represent 50 per cent of a Block of 20 verses but only 10 per cent of a Block of 100.
With percentages, two Blocks of different size can be placed on the same coordinate system and compared more proportionately.
Bazargan’s curve is therefore built from the pairs (k, pk), and not from (k, qk), which counts the tokens contained in each category. The curve shows the percentage of verses, and not the percentage of all tokens.
5.3. Joining the points
Once all the points have been placed on the coordinate plane, Bazargan joins them with a line:
«اگر آن نقاط را بهیکدیگر وصل کنیم یک خط منکسر مضرس بهدست میآید که آن را منحنی مشخصه آن سوره میگوییم.»
“If we join those points to one another, we obtain a broken, serrated line, which we call the characteristic curve of that surah.” —Sayr, I, pp. 56–57.
Formally, the original curve of a Block is the line joining all the available points (k, pk). Where no verse falls into a given length category, its frequency is zero and its percentage is zero as well.
The curve can therefore drop to the baseline and rise again at the next category. It is this that produces the pits, spikes, and serrated shape.
The word “curve” here should be understood as a visual convention. Token count is a discrete variable: a verse may be 4 or 5 tokens long, but not 4.5. The line between those two points does not indicate the existence of a verse 4.5 tokens long.
In modern statistical terminology the form is better described as a relative frequency polygon. Bazargan continues to call it a curve because his concern is primarily with the overall geometrical shape of the distribution.
5.4. A simple example: al-Humazah
The distribution of al-Humazah, on Bazargan’s counting, is:
| Length | Frequency | Percentage of verses |
|---|---|---|
| 3 tokens | 3 | 33.3% |
| 4 tokens | 5 | 55.6% |
| 5 tokens | 1 | 11.1% |
The points of the curve are (3, 33.3), (4, 55.6), and (5, 11.1). The curve rises from the 3-token category to a peak at 4 tokens, then falls at the 5-token category.
That peak shows the 4-token category to be the length most frequently found, so the dominant length is 4. Since five of the nine verses fall into that category, the height of the peak of the original curve is 55.6 per cent.
The curve of al-Humazah appears narrow and tall because most of its verses are concentrated within a very small range, namely 3 to 5 tokens.
5.5. What can be read from the shape of the curve
The characteristic curve allows several properties of the distribution to be seen at once.
Horizontal position
The position of the curve shows the region of lengths in which most verses are gathered. A curve displaced to the right has more long verses, while a curve concentrated on the left is dominated by short ones.
The horizontal position of the curve is nevertheless not identical with the mean length. A few very long verses can raise the mean without appreciably moving the peak of the distribution.
The position of the peak
The position of the peak of the original curve indicates the dominant length or mode. If the peak lies at 8 tokens, the most frequently found length category is 8 tokens.
If two categories share the maximum frequency, the distribution has two modes. Bazargan notes that some surahs may have two dominant lengths.
The height of the peak
The height of the peak shows the degree to which the verses are concentrated in a particular category. A tall, sharp curve means that most verses are gathered in one or a few adjacent length categories, while a low curve means they are spread across more.
The height of the peak therefore measures the concentration of the distribution, and not verse length.
The width of the curve
The horizontal width shows the diversity of verse lengths. A narrow curve means the lengths are relatively uniform, while a wide curve indicates greater variation.
Bazargan describes the curve of al-Humazah as tall and sharp, and that of al-Māʾidah as low and spread out. The curve of Ṭā Hā lies between the two.
Right and left tails
The portion of the curve after the peak is called the tail. If the curve extends far to the right, there are verses considerably longer than the dominant length.
A long right-hand tail can make the mean length exceed the mode. Conversely, where the curve spreads more to the left, the mean may fall below the dominant length.
Spikes and pits
The original curve may rise and fall sharply because some categories contain many verses while the neighbouring ones contain few or none at all.
Such spikes may arise from:
- a small Block size;
- natural variation in verse length;
- one or two outlying verses;
- the effects of the tokenisation rules;
- or the possibility that the Block still mixes several populations of verses.
A serrated curve does not by itself establish that a Block ought to be divided.
5.6. Comparing curves across Blocks
The curves of several Blocks may be placed on the same coordinate system as long as:
- the horizontal axis uses the same token unit;
- the vertical axis uses the percentage of verses;
- and the tokenisation rules are consistent.
Bazargan compares al-Humazah, Ṭā Hā, and al-Māʾidah in a single figure. On his observation, the curve of al-Humazah lies to the left and is tall and narrow; the curve of Ṭā Hā has a transitional position and shape; and the curve of al-Māʾidah is displaced to the right, lower, and wider.
From that comparison Bazargan proposes a general pattern: as revelation proceeds, the position of the curve shifts to the right, the dominant length increases, the peak tends to fall, and the distribution tends to widen.
Observation and chronological interpretation must still be kept apart here. The observation is that the curves of several surahs differ in position and shape. The conclusion that this difference represents a temporal development is Bazargan’s hypothesis, still to be tested across the whole corpus.
The curve itself carries no date.
5.7. The original characteristic curve
The curve formed directly from the empirical points is called the original characteristic curve. It has several properties:
- every point can be traced back to the distribution table;
- its shape is determined entirely by the data;
- categories with no verses produce zero points;
- it may be heavily serrated;
- and its results are reproducible as long as the data and the tokenisation rules are the same.
The original curve must be treated as the primary representation, since it preserves a direct relation to the frequencies actually observed.
In Bazargan’s indicator diagram the original curve is labelled ABC, with its peak at point B. The horizontal position of B indicates the dominant length, while its vertical position indicates the percentage of verses in the dominant category of the original distribution.
5.8. Why Bazargan drew an equivalent curve
The original curve is often serrated in shape. A small change in one or two verses can produce a spike or a pit that looks large visually, especially in Blocks of small size.
Bazargan therefore draws a second curve, called the equivalent characteristic curve (monḥanī-ye moshakhkhaṣeh-ye moʿādel), with the aim of displaying the general tendency of the distribution while reducing the effect of fluctuation in individual length categories.
He explains:
«برای آنکه از اثر تضاریس کاسته شود … بهجای منحنی مضرس حفرهدار، یک منحنی منظمی که معادل آن باشد رسم کنیم.»
“In order to reduce the effect of the ruggedness and obtain a simpler form, in place of the serrated and pitted curve we may draw a regular curve equivalent to it.” —Sayr, I, p. 59.
The equivalent curve is the product of smoothing the original curve. It is no new distribution and reflects no different counting of tokens.
5.9. How the equivalent curve is drawn
Bazargan gives no equation or fully mechanical smoothing algorithm. He describes the procedure graphically:
«منحنی منظم معادل تقریباً از وسط قطعات عبور داده و طوری کشیده میشود که هرجا در بالای منحنی مضرس اصلی میافتد در نقطه بعد در زیر قرار گیرد، بهنحوی که سطوح مثبت و منفی حاصل در بین دو منحنی یکدیگر را جبران کنند و مساحت زیر هر دو منحنی مساوی درآید.»
“The regular equivalent curve is passed roughly through the middle of the segments and is drawn so that wherever it falls above the original serrated curve, at the next point it lies below it, in such a way that the positive and negative areas between the two curves compensate each other and the area beneath both becomes equal.” —Sayr, I, p. 60.
The equivalent curve must therefore satisfy three basic ideas:
- it follows the general tendency of the original curve;
- it balances the spikes and the hollows;
- it preserves an overall area equivalent to that of the original curve.
The third principle amounts to saying that the area beneath the original curve and the area beneath the equivalent curve are approximately equal. Expressing this with integral notation only states the principle of equal areas; Bazargan himself performed the smoothing graphically, without any explicitly formulated continuous function.
He generally draws the equivalent curve with a single peak, so that its shape resembles a hood or a cap. He names Surah al-Jumuʿah as an exception, because its distribution yields two peaks.
5.10. An example: the equivalent curve of al-Anbiyāʾ
Bazargan uses al-Anbiyāʾ as his principal example of smoothing. Its original curve has several spikes:
- rising towards a main concentration at around 8 tokens;
- falling sharply at the next category;
- rising again at another category;
- then declining irregularly towards the right.
The equivalent curve is drawn through the middle of those fluctuations. It preserves the general direction — rising to a peak, falling gradually, then approaching zero — without following every rise and fall.
The example brings out the different functions of the two curves:
| Curve | Question answered |
|---|---|
| Original curve | What is the empirical frequency of each length category? |
| Equivalent curve | What is the tendency of the shape once local fluctuation has been smoothed away? |
The equivalent curve does not replace the original. The two carry different information.
5.11. The curve and Bazargan’s four indicators
The curve is then used together with the distribution table to obtain several indicators. Not all the indicators come from the same curve.
| Indicator | Source |
|---|---|
| Mean length | Total tokens divided by verse count |
| Dominant length L | The mode of the distribution table or the peak of the original curve |
| Height H | The maximum height of the equivalent curve |
| Dāmaneh | A combined measure of the widening and asymmetry of the curve |
In Bazargan’s diagram, ABC is the original curve and AbC the equivalent curve; B is the peak of the original and b the peak of the equivalent; OX indicates the dominant length and OY the maximum height of the equivalent curve.
This distinction matters. The height Bazargan uses as a formal indicator is the height of the peak of the smoothed curve, and is not always the same as the highest percentage in the table.
Forming the curve is therefore an intermediate stage between the distribution and the computation of the indicators: the distribution table yields both the original and the equivalent curve, the two yield the indicators, and it is the indicators that enter ṭūl-e mabnā.
5.12. The curve does not record the order of verses
The curve preserves only the relation between length and frequency, and shows nothing of the position of verses within a Block. The series 3, 3, 4, 4, 5 and the series 5, 4, 3, 4, 3 produce the same curve, since their length frequencies are identical.
The curve therefore cannot show:
- whether the long verses stand at the beginning or the end of a Block;
- whether verse length changes gradually;
- changes of theme and rhyme;
- narrative continuity;
- or the position of outlying verses.
The curve consequently cannot replace textual reading and the process of segmentation.
5.13. The limitations of the equivalent curve
The equivalent curve is one of the hardest parts of Bazargan’s method to reproduce, since it is not determined by any explicit algorithm.
Two researchers may draw slightly different equivalent curves even from the same distribution table. Such differences in drawing can alter:
- the position of the peak;
- the maximum height;
- the assessment of width and asymmetry;
- and the indicators subsequently used in the ṭūl-e mabnā calculation.
Smoothing also risks concealing important information. A second spike may be treated as random variation when it in fact points to the presence of two distinct populations of verses.
In MushafNuzuli’s digital reconstruction, therefore:
- the original curve must be formed directly from the data and always displayed;
- the equivalent curve must be stored as derived data;
- the smoothing method must be documented;
- the equivalent curve must not alter the original distribution table;
- indicators derived from the equivalent curve must be distinguished from those computed directly;
- and the digital results must not be given an air of greater objectivity than Bazargan’s actual procedure possessed.
Where Bazargan’s equivalent curve is reproduced by a modern algorithm, the result must be called an algorithmic reconstruction, unless it has genuinely been digitised from his own drawing.
5.14. The place of the curve within the method as a whole
The curve serves three main functions in Bazargan’s method.
First, it condenses a frequency table into a form that is easy to compare visually.
Second, it displays at once the dominant position, the concentration, the widening, the tails, and the irregularity of the distribution.
Third, the equivalent curve provides the basis for the height and dāmaneh indicators, which are then combined with mean length and dominant length.
The curve itself does not yet determine chronology. It produces a statistical description that will be translated into a comparative measure at the next stage.
The most accurate formulation is:
The characteristic curve is a geometrical representation of the distribution of verse lengths within a Block, formed by placing token counts on the horizontal axis and the percentage of verses on the vertical. The original curve preserves the empirical frequencies, while the equivalent curve smooths their fluctuation in order to obtain the general shape of the distribution.
The methodological sequence becomes:
tokenisation → verse lengths → segmentation → Blocks → distribution → original and equivalent curves → curve indicators → ṭūl-e mabnā
The stage after the curve is to explain how Bazargan obtains his four indicators, and why only three of them finally enter ṭūl-e mabnā.
6. The four indicators and the composite length index (ṭūl-e mabnā)
Having formed the original characteristic curve and the equivalent curve, Bazargan summarises the distribution of verse lengths in four indicators. Each is designed to capture a different aspect of the distribution:
- the mean length;
- the dominant length;
- the height of the equivalent curve;
- and the dāmaneh.
Bazargan calls the combination of these indicators طول مبنا (ṭūl-e mabnā). Literally the term may be rendered “reference length” or “base length”, though neither translation says much about how the value is formed.
Bazargan himself supplies an explanatory equivalent:
طول متوسط تصحیحشده “the corrected mean length.”
The value of ṭūl-e mabnā is formed by normalising and averaging several distinct indicators, which makes it something other than an arithmetic mean of verse lengths corrected by a conventional statistical procedure. In this reconstruction the term used is therefore:
the composite length index (ṭūl-e mabnā)
The word “index” makes clear that the figure is a composite score and does not report an observed verse length. The Persian term is retained because the procedure is a construction peculiar to Bazargan with no exact counterpart in modern statistical terminology.
6.1. Why the mean length is not considered sufficient
The simplest measure for a Block is the mean verse length, that is, the total token count W divided by the verse count N.
The difficulty is that a mean can be pulled about by a small number of very long verses. A Block whose verses are mostly short may acquire a relatively high mean simply because it contains one or two long ones.
The series 4, 4, 4, 4, 14, for instance, yields a mean of 6, although four of its five verses are 4 tokens long. A mean of 6 fails to indicate the main point of concentration in that distribution.
Bazargan therefore refused to rely on a single measure. He sought to combine several characteristics of the distribution so that a deviation in one indicator might be offset by another — moving from a single, easily distorted indicator, to several indicators, and then to a more stable composite index.
6.2. The first indicator: mean length
The first indicator is the mean verse length, W divided by N. This measure takes account of every verse and every token in the Block, and answers the question:
If all the tokens were distributed evenly among all the verses, how many tokens would each verse receive?
The mean length is a statistic computed directly from the data, independent of how the equivalent curve is drawn. Its strengths:
- it uses every observation;
- it is easy to compute;
- it can be reproduced exactly;
- and it is sensitive to changes in the overall length of the text.
Its weakness is sensitivity to outliers. A few very long verses can raise the mean even though the majority of verses remain short.
6.3. The second indicator: dominant length
The second indicator is the dominant length (ṭūl-e ghāleb), the length category with the highest verse frequency. Statistically this is the mode, written L. If the 8-token category has the largest number of verses, L equals 8.
On the original curve the dominant length is given by the horizontal position of the empirical peak, labelled OX in Bazargan’s diagram.
The dominant length answers the question:
What verse length in tokens is most frequently found in this Block?
The mode is not much affected by a single very long verse, and in that respect it is sturdier than the mean. Its weakness lies elsewhere:
- it can shift as a result of a change in one or two verses;
- it can take two values where the distribution is bimodal;
- and it makes no use of the information in the remaining length categories.
The dominant length therefore complements the mean length without replacing it. The two capture different properties of the distribution.
6.4. The third indicator: the height of the equivalent curve
The third indicator is the maximum height of the equivalent characteristic curve, written H and labelled OY in Bazargan’s diagram. The value of H gives the percentage of verses at the peak of the distribution once the original curve has been smoothed.
The taller the curve, the more the verses are concentrated within a particular range of lengths. A low curve indicates that the verses are spread across more length categories. A high value of H thus marks a more concentrated distribution, and a low value a more dispersed one.
Bazargan observed that the early surahs tend to have tall, narrow curves, while the groups he placed later tend to have lower and wider ones.
Because the direction of change in height runs opposite to that of the two preceding indicators, Bazargan uses the reciprocal of the height when normalising. Mean length and dominant length are expected to rise over time, while the height of the curve is expected to fall.
It should be stressed that H derives from the equivalent curve, and is not always the same as the highest percentage in the empirical table. The two may coincide, but they may also differ as a result of the smoothing.
6.5. The fourth indicator: dāmaneh
The fourth indicator is called دامنه (dāmaneh), which may literally mean range, span, or amplitude. In Bazargan’s method it amounts to more than the difference between the maximum and minimum lengths.
He formulates it as the square root of the product of two ratios: XC divided by AM, multiplied by 100 divided by H. In his diagram, A is the left-hand limit of the curve, C the right-hand limit, X the position of the dominant length, M the position of the mean length, and H the height of the equivalent curve.
The ratio of XC to AM captures the imbalance in the span of the curve to the right and to the left. The factor of 100 divided by H brings in the effect of the curve’s height: two curves may have a similar horizontal span, yet the lower one counts as more dispersed.
The square root is used so that the result becomes the geometric mean of the two components — asymmetry multiplied by vertical widening. Dāmaneh is thus a combined index of the widening and asymmetry of the curve, and not a simple range between the largest and smallest lengths.
6.6. The direction of change in the four indicators
On Bazargan’s hypothesis the four indicators are related to the progress of revelation, though not all in the same direction.
| Indicator | Symbol | Direction assumed by Bazargan |
|---|---|---|
| Mean length | — | Rising |
| Dominant length | L | Rising |
| Height of the equivalent curve | H | Falling |
| Dāmaneh | D | Rising |
Because H moves downwards, Bazargan inverts it when forming the relative indicators. Under that transformation the mean length, the dominant length, the reciprocal of the height, and the dāmaneh are all expected to move upwards. The step is necessary if the indicators are to be averaged without cancelling one another out.
6.7. Why the indicators cannot simply be averaged
The four indicators have different units and scales:
- mean length is expressed in tokens per verse;
- dominant length is expressed in tokens;
- height is expressed as a percentage;
- dāmaneh is a derived index.
Adding the four together and dividing by four is therefore meaningless — a percentage cannot simply be added to a token count. All the indicators must first be converted to a comparable scale. In modern statistics this step is called normalisation against a reference value.
6.8. Surah al-Anbiyāʾ as the reference Block
Bazargan chose Surah al-Anbiyāʾ as the basis of normalisation. On his calculations the reference values used are:
| Indicator | Reference value for al-Anbiyāʾ |
|---|---|
| Mean length | 10 |
| Dominant length | 8 |
| Height of the equivalent curve | 15% |
| Dāmaneh | 3.66 |
Three considerations lay behind the choice of al-Anbiyāʾ.
First, its mean length is close to 10, which makes hand computation easier.
Second, its characteristic curve was regarded as sufficiently regular, and its equivalent curve lies relatively close to the original.
Third, al-Anbiyāʾ is placed in the middle region of the development of verse length, so that the relative values of other surahs become neither too large nor too small.
Al-Anbiyāʾ here serves as a numerical reference Block for putting the indicators on a common scale, and not as an absolute theological or chronological standard.
6.9. Forming the relative indicators
Each indicator of a Block is compared with the corresponding indicator of al-Anbiyāʾ:
| Relative indicator | How it is computed |
|---|---|
| r for mean length | mean length divided by 10 |
| rL | L divided by 8 |
| rH | 15 divided by H |
| rD | D divided by 3.66 |
The division for height is deliberately inverted, since the height is assumed to fall. Where H is written as a percentage, the figure used is the percentage number — for a height of 30 per cent, for instance, rH is 15 divided by 30, that is 0.5.
All these results are dimensionless and represent ratios to the reference Block. Where an indicator equals the value for al-Anbiyāʾ, its relative value is 1. A value below one shows that the indicator lies below the reference; a value above one shows that it exceeds it.
6.10. From four indicators to three principal ones
At the outset Bazargan computed the average of all four relative indicators. After comparing more surahs, he found that the dāmaneh often deviated widely from the other three.
The main cause is the sensitivity of the dāmaneh to long verses in the tail of the distribution. The presence or absence of a single long verse can move the right-hand limit of the curve, changing the distance XC considerably. The value of D may then shift sharply while the mean length, the mode, and the height of the curve change only slightly.
Bazargan therefore removed the dāmaneh from the final calculation. The indicator continues to be computed and is used as a diagnostic or supporting measure.
The final distinction becomes:
| Standing | Indicator |
|---|---|
| Principal indicators | Mean length, dominant length, and height |
| Supporting indicator | Dāmaneh |
Although the discussion begins with four indicators, the final version of ṭūl-e mabnā is therefore built from three principal ones.
6.11. The formula for the composite length index
The average of the three relative indicators is obtained by adding the r for mean length, rL, and rH, and dividing by three. Since the result is a dimensionless ratio, Bazargan multiplies it by 10, the mean length of the reference Block.
Bringing in all the indicators, the index T is obtained as follows:
T = 10 × [ (mean length ÷ 10) + (L ÷ 8) + (15 ÷ H) ] ÷ 3
It is this value T that is called ṭūl-e mabnā, or in the terminology of this reconstruction, the composite length index.
The formula can be simplified to:
T = ( mean length + 1.25 L + 150 ÷ H ) ÷ 3
The first form is better retained in a methodological account, since it displays the normalisation against al-Anbiyāʾ more clearly.
6.12. Why the result is multiplied by 10 again
The average of the three relative indicators is a dimensionless figure, say 0.383. Such a number no longer looks like a measure of verse length. Bazargan multiplies it by 10 so that the final result returns to a scale resembling mean length, turning 0.383 into 3.83.
The step changes only the scale of presentation and leaves the order of the Blocks untouched. Ṭūl-e mabnā should therefore be understood as a score on a length-equivalent scale, and not as an actual token count.
6.13. A worked example: al-Humazah
For al-Humazah, on Bazargan’s calculations, the mean length is 3.78, the dominant length 4, and the height of the equivalent curve 55.5 per cent.
| Relative indicator | Calculation | Result |
|---|---|---|
| Mean length | 3.78 ÷ 10 | 0.378 |
| Dominant length | 4 ÷ 8 | 0.500 |
| Height (inverted) | 15 ÷ 55.5 | ≈ 0.270 |
The average of the three is (0.378 + 0.500 + 0.270) ÷ 3, that is roughly 0.383. Returned to the length scale by multiplying by 10, this gives a ṭūl-e mabnā for al-Humazah of approximately 3.83.
The value is close to its mean length of 3.78 without the two being identical. The difference comes from two further pieces of information: the dominant length of 4 tokens, and the concentration of the distribution as reflected in the height of the curve.
6.14. A control test on al-Anbiyāʾ
Since al-Anbiyāʾ is the reference Block, all its relative indicators equal one: 10 divided by 10, 8 divided by 8, and 15 divided by 15. Their average is 1, and multiplied by 10 this gives a ṭūl-e mabnā for al-Anbiyāʾ of exactly 10.
The result shows that the scale of ṭūl-e mabnā is indeed anchored to the mean length of al-Anbiyāʾ. This is a mathematical consequence of the normalisation procedure, and not an independent finding about al-Anbiyāʾ.
6.15. What ṭūl-e mabnā actually measures
Ṭūl-e mabnā combines three properties of the distribution: overall length, the dominant centre, and the degree of dispersion.
| Component | Information carried |
|---|---|
| Mean length | The overall length of the verses |
| Dominant length L | The most common length category |
| Reciprocal of the height (1 ÷ H) | The degree of widening, or the lowness of the curve’s concentration |
The value of ṭūl-e mabnā rises when the mean length rises, when the dominant length shifts to the right, or when the equivalent curve becomes lower. Two Blocks with the same mean length may therefore have different values of ṭūl-e mabnā if their modes or degrees of concentration differ.
6.16. Ṭūl-e mabnā is not an ordinary mean length
The conceptual difference may be summarised as follows:
| Measure | How it is formed | Meaning |
|---|---|---|
| Mean length | Total tokens divided by verse count | Empirical mean |
| Dominant length | The category with the highest frequency | Empirical mode |
| Ṭūl-e mabnā | The average of three normalised indicators | Composite index |
A statement such as “the mean verse length of this Block is 12.4 tokens” is therefore mistaken if the figure 12.4 is in fact a ṭūl-e mabnā. The accurate formulation is:
“This Block has a composite length index (ṭūl-e mabnā) of 12.4.”
The value lies on a scale resembling a token count, without always equalling the observed mean length.
6.17. Can the term “corrected mean length” be used?
The term may be used when translating Bazargan’s own explanation directly, since he himself writes of “the corrected mean length, or ṭūl-e mabnā”.
In modern statistical terminology, however, “corrected mean” can suggest that Bazargan corrected estimator bias, applied a sample-size correction, removed outliers by a formal rule, or computed an adjusted mean from a regression model. He did none of these things.
What Bazargan did was to normalise, invert the direction of the height, average three ratios, and rescale. “Composite length index” is therefore the more transparent statistical description, with the term ṭūl-e mabnā retained so that the link to the original terminology is not lost.
6.18. The hidden weighting assumption
Bazargan averages the three relative indicators with equal weight, one third each. Mean length, dominant length, and the reciprocal of the height are thus treated as equally important once normalised.
That weighting is a methodological decision of Bazargan’s, and not an estimate derived statistically from the corpus. This does not in itself invalidate the index, but it must be stated, so that readers do not suppose the formula was arrived at through optimisation or the testing of a statistical model.
6.19. The limits of normalising against al-Anbiyāʾ
Normalising against al-Anbiyāʾ makes different indicators comparable. The reference values themselves, however, depend on a number of earlier decisions:
- the tokenisation rules;
- the Block boundaries;
- the distribution of verse lengths;
- the drawing of the equivalent curve;
- the determination of the dominant length;
- and the measurement of the curve’s height.
If MushafNuzuli uses tokenisation rules different from Bazargan’s, the reference values of 10, 8, and 15 per cent must not be taken as fixed. The indicators for al-Anbiyāʾ need to be recomputed under the same rules as every other Block.
The underlying principle is simple: the rules for measuring a Block must be exactly the rules for measuring the reference Block. Without that consistency, normalisation loses its meaning.
6.20. Sources of uncertainty
Not all the components of ṭūl-e mabnā are equally reproducible.
| Component | Reproducibility |
|---|---|
| Mean length | High; computed directly |
| Dominant length | High, except in multimodal distributions |
| Height of the equivalent curve | Lower; depends on the smoothing |
| Dāmaneh | Lower; depends on the curve and the tail of the distribution |
The greatest uncertainty lies in H. Because Bazargan’s equivalent curve was drawn graphically, two reconstructions may yield slightly different heights. A change in height alters the value of 15 divided by H, and hence the ṭūl-e mabnā.
A digital reconstruction therefore needs to store separately:
- the empirical height of the curve;
- the height of Bazargan’s equivalent curve;
- the height produced by algorithmic smoothing;
- the smoothing method used;
- and the value of ṭūl-e mabnā yielded by each method.
6.21. The place of the index in determining chronology
Ṭūl-e mabnā is not yet a date of revelation. It yields a single comparative score for each Block, and the Blocks can then be ordered from the lowest score to the highest.
Bazargan interprets a rising value as development towards a later phase of revelation. That relation remains a hypothesis to be tested, and ṭūl-e mabnā is not identical with the year of revelation.
A change in the index shows a statistical difference in the length and distribution of verses. Translating that difference into an order and a year of revelation is the next methodological stage.
6.22. Terminological usage in MushafNuzuli
To preserve both accuracy and readability, the following terms may be used consistently:
| Context | Term used |
|---|---|
| Principal name | Composite length index |
| Source term | Ṭūl-e mabnā |
| Bazargan’s own gloss | Corrected mean length |
| Short label in tables | Ṭūl-e mabnā |
| Tooltip definition | A composite index of mean length, dominant length, and curve height, normalised against al-Anbiyāʾ |
At first occurrence the term may be written as the composite length index (ṭūl-e mabnā), which Bazargan also calls the “corrected mean length”. Thereafter ṭūl-e mabnā alone will serve.
In this way the Persian term is preserved without leaving readers to take it literally as “base length”.
6.23. The result of the indicator stage
For each Block this stage produces a set of values comprising the mean length, the dominant length, the height, the dāmaneh, their four relative indicators, and the index T. The principal value is T, the ṭūl-e mabnā, while the dāmaneh is retained as a diagnostic indicator.
The full sequence becomes:
length distribution → original curve → equivalent curve → four indicators (mean length, dominant length, height, dāmaneh) → normalisation → composite length index
Ṭūl-e mabnā may therefore be formulated as:
A composite index on a length-equivalent scale, formed by averaging the mean length, the dominant length, and the reciprocal of the curve height, after all three have been normalised against the reference values of Surah al-Anbiyāʾ.
The index is designed to yield a measure more stable than mean length alone. It nevertheless remains dependent on the choice of reference Block, on the weighting Bazargan fixed, and above all on the drawing of the equivalent curve.
6.24. The relative order of the Blocks
By relative order is meant the position of a Block compared with the other Blocks on the basis of its ṭūl-e mabnā. At this stage Bazargan has not yet fixed any date or year of revelation.
Suppose the following values are obtained:
| Block | Ṭūl-e mabnā |
|---|---|
| A | 3.8 |
| B | 5.2 |
| C | 7.1 |
| D | 10.4 |
If it is assumed that the length and dispersion of verses increase over the course of revelation, the relative order is A, B, C, D — with A estimated earlier than B, B earlier than C, and C earlier than D.
That order does not yet tell us:
- how many years separate A from B;
- whether B and C were revealed in the same year;
- the exact date of any Block;
- or its position in the canonical order of the muṣḥaf.
The only relation that holds is that a lower value of T marks a Block estimated to be earlier. The relation is ordinal, not temporally quantitative. A difference of 2 in ṭūl-e mabnā does not mean a difference of two years, two months, or two stages of revelation.
This must also be distinguished from the normalisation against al-Anbiyāʾ. Al-Anbiyāʾ is used to put the indicators on a common scale and is fixed at a ṭūl-e mabnā of 10. Relative order is obtained by comparing all the ṭūl-e mabnā values with one another, and not merely by determining whether a value is greater or smaller than that of al-Anbiyāʾ.
Stated more fully: once every Block has been given a ṭūl-e mabnā, the Blocks are ordered from the lowest value to the highest. The result is a relative order, a statistical ranking showing which Blocks are estimated to be earlier or later than others. To convert it into an estimated year or phase of revelation, Bazargan requires a further stage of chronological calibration.
One important limitation applies. Where two Blocks have very close values — 7.21 and 7.26, say — their order must not be treated as settled. So small a difference may be smaller than the uncertainty introduced by the smoothing of the equivalent curve. A digital presentation should therefore allow three categories: earlier, later, or not confidently distinguishable.
In brief:
Relative order is a ranking of Blocks by ṭūl-e mabnā. It is neither a date of revelation nor a measure of the interval of time between those Blocks.
The next stage is to use that order to estimate the year of revelation.
7. Historical calibration: from relative order to an estimated year of revelation
Once all the Blocks have been ordered by ṭūl-e mabnā, Bazargan has a relative order of revelation. That order shows that one Block is estimated to be earlier or later than another, and stops there. It does not yet answer:
- in which year a given Block was revealed;
- whether a Block belongs to the Meccan or the Medinan period;
- how much time separates two Blocks;
- and which part coincides with a particular event in the life of the Prophet.
To answer these questions Bazargan adds a further stage, called in this reconstruction historical calibration — the process of linking the statistical order to a scale of time through a number of verses or Blocks whose dates can be estimated from historical sources.
In brief: order by ṭūl-e mabnā → cumulative token count → historical anchor points → calibration curve → estimated year of revelation.
7.1. Ṭūl-e mabnā is not converted directly into a year
The first point to stress is that Bazargan draws no direct equation between ṭūl-e mabnā and the year of revelation. A difference of two units in ṭūl-e mabnā does not mean a difference of two years. The function of the index is confined to establishing an order: a smaller value of T marks a Block estimated to be earlier.
Once that order is established, Bazargan turns to a different variable, namely the token count of each Block. It is the token count that is then used to determine a Block’s cumulative position within the whole process of revelation.
Two distinct stages are involved:
| Stage | Principal variable | Result |
|---|---|---|
| Ordering | Ṭūl-e mabnā | Relative order of the Blocks |
| Calibration | Cumulative tokens and historical dates | Estimated year of revelation |
The distinction matters. Ṭūl-e mabnā fixes a Block’s position in the order, and is not itself a measure of time.
7.2. Computing cumulative tokens
Suppose each of the ordered Blocks has a given token count. Bazargan then computes the number of tokens accumulated up to each Block, that is, the sum of the tokens of all preceding Blocks together with the Block itself.
For example:
| Block order | Tokens in the Block | Cumulative tokens |
|---|---|---|
| 1 | 120 | 120 |
| 2 | 180 | 300 |
| 3 | 250 | 550 |
| 4 | 310 | 860 |
The fourth Block stands at cumulative position 860. That figure means that, on Bazargan’s ordering, some 860 tokens had been revealed from the beginning up to the completion of the fourth Block.
This cumulative count turns the list of Blocks into a steadily rising trajectory. That trajectory still has no unit of years, and to place it on a timeline a number of historical anchor points are required.
7.3. Historical anchor points
Several parts of the Qur’an are linked by the sīrah sources, by asbāb al-nuzūl reports, or by their own content to events whose time can be estimated. Bazargan calls these points گذرگاههای تنزیل (gozargāh-hā-ye tanzīl), literally “waypoints of revelation”.
Examples include Blocks associated with:
- the beginning of revelation;
- the Battle of Badr;
- the Battle of Uḥud;
- the affair of al-ifk;
- several other Medinan events;
- and passages placed towards the end of the Prophet’s life.
An anchor point carries two pieces of information: an estimated time for the event, and the cumulative token position of the Block associated with it. As a hypothetical example, if a Block associated with an event in the second year of the Hijra stands at a cumulative position of 50,000 tokens, the point is the pair (2 AH, 50,000).
Such a point ties a particular position in the statistical order to a particular position in historical time.
7.4. Constructing the calibration plane
Bazargan places these historical points on a Cartesian coordinate system. The horizontal axis shows time, or the year of the prophetic mission, while the vertical axis shows the cumulative token count revealed. Each Block linked to a historical event yields one point on that plane.
In his account, Bazargan states that these points turned out to lie fairly close to a straight line. He calls the resulting graph the منحنی تنزیل (monḥanī-ye tanzīl), or “curve of revelation”. He describes the procedure as joining the year in which an event took place with the number of tokens that, on the statistical ordering, had accumulated by that time. Bazargan, “The Quantity of Verses as a Measure of the Course of Revelation,” pp. 75–78.
7.5. The linear curve of revelation
Bazargan reads the closeness of the historical points to a straight line as a sign that the cumulative token count grows at a relatively constant annual rate.
In general form, the line states that the cumulative token count at a given time equals a starting position plus an annual rate multiplied by time. The slope of the line is the change in cumulative tokens divided by the change in time, and that value gives the number of tokens which, on the model, accumulate on average in one year.
Bazargan arrives at a rate usually summarised as about 3,630 tokens per year.
This figure is no universal linguistic constant. It is the product of:
- Bazargan’s tokenisation rules;
- the division of the Qur’an into Blocks;
- the ordering of the Blocks by ṭūl-e mabnā;
- the choice of historical anchor points;
- and the line drawn through those points.
The figure of 3,630 must therefore be described as an estimate within Bazargan’s model, and not as a fact standing apart from the procedure that produced it.
7.6. Did Bazargan perform a linear regression?
In modern statistical terminology, Bazargan’s procedure resembles the construction of a linear calibration model with a term for the deviation between the historical points and the fitted line.
Bazargan himself reports no regression estimates in the modern statistical sense. He constructed his curve graphically on Cartesian coordinates and observed that the historical points lay roughly on a single straight line.
The more accurate term for his original procedure is therefore:
graphical linear calibration
Where MushafNuzuli recomputes that line using least squares, robust regression, or a probabilistic model, the result must be called:
an algorithmic reconstruction of Bazargan’s calibration curve
The results of a modern regression must not simply be equated with the line Bazargan drew.
7.7. Inverting the equation to obtain a time
Once the relation between cumulative tokens and time has been established, the equation can be inverted: the estimated time is obtained by subtracting the starting position from the cumulative token count and dividing by the annual rate. Each Block thus acquires a temporal position — from Block, to cumulative tokens, to estimated year.
Using the simple form with a rate of 3,630 tokens per year, the end of the first year falls at around 3,630 tokens, the end of the second at around 7,260, and the end of the third at around 10,890.
Blocks lying between 3,630 and 7,260 cumulative tokens are placed roughly in the second year, while those between 7,260 and 10,890 fall roughly in the third.
The actual boundary figures follow the line and the division of years that Bazargan used, including the fact that the first and last years of revelation are not full years.
7.8. Blocks that cross a year boundary
A Block does not always end exactly on an annual boundary.
Suppose a Block begins when the cumulative token count stands at 6,900 and ends at 7,500. If the boundary between the second and third years lies at 7,260, that Block crosses the year boundary — on the model it begins at the end of the second year and finishes at the beginning of the third.
Where a system offers only a single year label, such a Block may have to be assigned to one year by some rule. A more accurate digital presentation can store its range instead, namely the estimated time at the start of the Block and the estimated time at its end, so that a Block is not always forced to carry a single date.
7.9. The role of historical events within the method
Historical events may serve two distinct functions, which must be kept apart.
| Function | Explanation |
|---|---|
| Calibration | The event is used to establish the relation between cumulative tokens and years |
| Validation | The event is used to test a chronology built without using it |
If an event has been used to draw the calibration line, the agreement of that point with the line cannot at the same time count as independent confirmation of the line.
In a digital reconstruction, the historical points should be divided into:
- calibration points, used to form the line;
- validation points, historical evidence held back and used only to test the result;
- contested points, events whose date or connection to a verse remains disputed.
That separation prevents circular reasoning, since the data forming the model must not also be the data testing it.
Behnam Sadeghi notes that Bazargan examined fifteen passages carrying historical information and that thirteen of them followed the expected order. Sadeghi also stresses that Bazargan’s chronological results are statistical in character and must not be read with greater precision than the data can bear. Sadeghi, “The Chronology of the Qurān,” Arabica 58 (2011): 210–299.
7.10. Calibration is not historical proof
Historical calibration does not prove that every link between a verse and a historical event is correct. It uses those links as connecting points between two scales, the cumulative token position and historical time.
If the dating of a point changes, the calibration line may change with it. Likewise, if a Block turns out not to be connected with the event assumed, that point must be removed or given lower weight.
The final result must therefore be phrased as:
“On Bazargan’s calibration, this Block is estimated to belong to a given year.”
A formulation such as “statistical calculation proves that this Block must have been revealed in that year” claims far more than the method can support.
7.11. Sources of uncertainty
There are at least five sources of uncertainty in this calibration.
First, the historical date of an event may be disputed.
Second, the connection between an event and a particular verse is not always certain. Asbāb al-nuzūl reports may differ or contradict one another.
Third, the cumulative position depends on the order of the Blocks. If two Blocks exchange places, the cumulative totals at every later point change as well.
Fourth, the token count depends on the tokenisation rules. Differences in counting particles, the basmalah, the disconnected letters, or repetitions can alter the slope of the line.
Fifth, the linear model assumes a relatively constant average rate. It cannot be excluded that the rate of revelation actually varied between periods, even if certain points appear to approach a straight line.
These five sources accumulate into an uncertainty in the calibrated year.
Bazargan himself did not treat the years he produced as absolute. Sadeghi indeed records Bazargan’s admission that the actual date of a passage might deviate by some two or three years from the year assigned by the model.
7.12. A year as an interval, not an exact date
Because of these uncertainties, the results of calibration are better not displayed as exact dates. Rather than writing “this Block was revealed in the fifth year”, it is more accurate to write:
This Block is placed by Bazargan’s model in the fifth year, with possible displacement to the neighbouring years.
In a digital reconstruction the result may be presented in three layers:
| Type of value | Example |
|---|---|
| Model position | Year 5.4 |
| Bazargan’s label | Year 5 |
| Uncertainty range | Roughly years 4 to 7 |
That range need not always be two or three years. Its width should be computed from the uncertainty of the historical points, the variation in the calibration line, the size of the Block, and the stability of its ordering.
7.13. Mission years are not Hijri years
Bazargan’s time axis covers the whole span of the prophetic mission, and is not confined to the Hijri years. Time labels therefore need to distinguish:
- years since the beginning of revelation, or the biʿthah;
- years before the Hijra;
- and years after the Hijra.
For instance, “the fifteenth year of the mission” is not the same as 15 AH. If the Meccan period is reckoned at about thirteen years, the fifteenth year of the mission falls roughly in the second year of the Hijra.
In a database it is best to store separately: mission_year, meccan_or_medinan, hijri_year, and, if required, the Gregorian equivalent. This separation prevents Bazargan’s year numbers from being misread as Hijri calendar dates.
7.14. What calibration actually produces
Historical calibration produces an estimate of a function converting cumulative token position into temporal position. Since the cumulative position can only be computed once the Blocks have been ordered, the whole process runs from Block, to ṭūl-e mabnā, to relative order, to cumulative tokens, and then to an estimated time.
The year of revelation is therefore not a direct product of verse length. It results from the combination of:
- the measurement of length;
- the ordering of the Blocks;
- the computation of cumulative tokens;
- the choice of historical anchor points;
- the construction of the calibration line;
- and the division of that line into yearly intervals.
7.15. Presentation in MushafNuzuli
For each Block, a digital reconstruction should store:
| Data | Description |
|---|---|
| Ṭūl-e mabnā | The index determining the order |
| Rank | The relative position of the Block |
| Token count | Tokens within the Block |
| Opening cumulative tokens | Position before the Block |
| Closing cumulative tokens | Position after the Block |
| Anchor status | Calibration, validation, or not an anchor |
| Historical source | The report or study underlying the date |
| Anchor date range | Where the dating is not a single value |
| Model time | Fractional year position |
| Bazargan’s year | The year label in his own table |
| Uncertainty | Range or degree of confidence |
| Calibration version | Bazargan’s original or an algorithmic reconstruction |
The site must also distinguish three kinds of result:
- Bazargan’s original values, the years given in his tables;
- reproduced results, the years obtained by following his procedure;
- modern reconstructions, estimates from a statistical model whose algorithm is stated explicitly.
The distinction is necessary because changes in tokenisation, segmentation, or the method of fitting the line can yield year boundaries that are not entirely the same.
7.16. The place of historical calibration within the method
Historical calibration is the bridge between statistical analysis and historical reconstruction. Before calibration, the relation between Blocks means only that one Block is estimated to be earlier than another. After calibration, each Block acquires an estimated position within the period of the mission.
This change does not remove the hypothetical character of the model. Calibration merely adds a scale of time to an order established beforehand.
The most accurate formulation is:
Historical calibration is the process of anchoring the relative order of the Blocks to a scale of time by linking the cumulative token count at a number of Blocks with events whose historical dates can be estimated. From those points Bazargan constructs a curve of revelation approximating a straight line, and then uses the cumulative token boundary of each year to estimate the year of revelation for all the Blocks.
The full sequence becomes:
segmentation → length distribution → ṭūl-e mabnā → relative order → cumulative tokens → historical points → calibration curve → estimated year
Relative order derives from ṭūl-e mabnā, while the estimated year derives from the cumulative token position on a historically calibrated curve.
The next stage is to see how the whole series of Blocks is gathered into Groups.
8. Chronological Groups: from 194 Blocks to 22 analytical clusters
The term chronological Group in MushafNuzuli refers to the combination of a number of Blocks adjacent in Bazargan’s chronological order. Groups are used to present the development of the text in larger and statistically more stable units.
The distinction between Block and Group needs to be made explicit, because Bazargan himself calls the 194 units produced by his segmentation “groups of contemporaneous verses”. In MushafNuzuli’s structure the term Block is used for those 194 units, while chronological Group is used for the 22 clusters formed by Behnam Sadeghi in combining Bazargan’s Blocks.
In brief:
| Level | Number | Function |
|---|---|---|
| Segment | 284 | A run of consecutive verses forming part of a Block |
| Block | 194 | The basic unit in Bazargan’s chronology |
| Chronological Group | 22 | A combination of adjacent Blocks for analysis |
| Year of revelation | A scale of about 23 years | Estimated time from historical calibration |
The hierarchy runs from Segments contained within Blocks, and Blocks contained within chronological Groups.
The year of revelation does not stand at the same hierarchical level. A year is a temporal attribute estimated once the Blocks have been placed on the curve of revelation.
8.1. Why Blocks are combined into Groups
Some of Bazargan’s Blocks are too small to be analysed stylometrically on their own. Word frequencies, morphemes, verse lengths, and other indicators of style shift readily as a result of small sample size.
To reduce that problem, Sadeghi combines Blocks that are:
- adjacent in Bazargan’s order;
- of relatively similar verse-length profile;
- and chronologically estimated to come from neighbouring periods.
A Group is thus a union of several Blocks. Group 1, for example, contains Blocks 2 to 16; Group 2 contains Blocks 17 to 34; Group 3 contains Block 1 together with Blocks 35 to 65; and Group 22 contains Blocks 184 to 194.
The combination yields longer stretches of text, so that stylistic tendencies can be measured more stably. Sadeghi explains that the formation of these 22 Groups was intended to simplify the graphs and to make the analysis of small Blocks possible. He relates the 22 Groups loosely to the span of 23 years of revelation that Bazargan used. Sadeghi, “The Chronology of the Qurān,” pp. 238–239.
8.2. A Group is an analytical unit
The boundary between two Groups is not a historically known event of revelation. Nor does it necessarily mark:
- the turn of a year;
- the move from Mecca to Medina;
- the occurrence of a particular event;
- or an abrupt change of theme.
A Group boundary is an analytical clustering boundary. Sadeghi chose it so that adjacent Blocks with similar verse-length profiles could be analysed together.
A chronological Group is therefore best understood as:
A cluster of several Blocks adjacent in Bazargan’s order, combined so as to represent a region of time more stably in statistical analysis.
The word “chronological” indicates that the clustering follows the direction of the proposed order of revelation. It promises no fixed historical date for each Group.
8.3. The correlation of Groups with the year of revelation
The 22 chronological Groups were made to follow the trajectory of revelation across roughly 23 years. There is therefore a positive correlation between Group number and time: lower-numbered Groups are estimated to be earlier than higher-numbered ones.
That relation holds only as a general tendency, or on average. It does not mean that every part of one Group must be earlier than every part of the next, and it is not a numerical equation between Group number and year number.
Consequently:
- Group 3 does not automatically mean the third year;
- Group 12 does not mean the year 12 AH;
- Group 22 does not mean the year 22 AH;
- and one Group need not equal one year of revelation.
The difference in the counts alone shows that the two scales are not identical: there are 22 Groups, while the period of revelation lasted about 23 years.
The correct correlation runs: as the Group number rises, the chronological position is generally later. The Group number itself remains no year number.
8.4. Why one Group is not one year
Group boundaries are drawn on the basis of Block order, sample size, and similarity of verse-length profile. Year boundaries derive from the historical calibration curve on the basis of cumulative token counts. The two are formed by different procedures:
| Boundary | Basis of formation |
|---|---|
| Group boundary | Combination of adjacent and relatively similar Blocks |
| Year boundary | Cumulative token position on the curve of revelation |
| Meccan–Medinan boundary | The Hijra and historical reconstruction |
A Group may consequently fall mainly within one model year, cross the boundary of two years, cover parts of several neighbouring years, or have an uncertain span of years.
The relation of a Group to time is best computed through its cumulative position, that is, from the cumulative token counts before and after the Group, translated into estimated opening and closing times by the calibration curve.
The correct result for a Group therefore reads “Group 8 occupies a region of time from around one year to the next on Bazargan’s model”, and not “Group 8 is the eighth year of revelation”.
8.5. Mission year numbers are not Hijri years
When Bazargan divides revelation into some 23 years, what he means are years since the beginning of the prophetic mission, and not 23 Hijri years. His timeline covers the Meccan period together with the Medinan.
The early years are counted before the Hijra. “The fifteenth year of the mission” therefore falls roughly at the beginning of the Medinan period and is not the same as 15 AH.
MushafNuzuli should distinguish:
| Label | Meaning |
|---|---|
| Group 15 | The fifteenth analytical cluster |
| Year 15 of the mission | A year counted from the beginning of revelation |
| 2 AH | The second year of the Hijri calendar |
| Medinan period | A historical category after the Hijra |
Sadeghi notes that Groups 12–22 generally correspond to the Medinan material in the traditional dating. This still does not make Group 12 identical with the year 12 AH; the Group number remains an analytical sequence number. Sadeghi, “The Chronology of the Qurān,” pp. 228–229.
8.6. The exception of Block 1 within Group 3
The structure of 22 Groups carries one important exception. The first five verses of Surah al-ʿAlaq, Q 96:1–5, are numbered Block 1 because Bazargan follows the report placing them as the first revelation.
On the statistical profile of verse length, their position falls later than a number of the early Blocks. In Sadeghi’s clustering, Block 1 is therefore placed in Group 3 together with Blocks 35 to 65.
Its Block identity number is retained as Block 1, while its analytical membership is placed in Group 3.
The case brings out the difference between:
- priority based on historical report;
- position based on statistical characteristics;
- the identity number of a Block;
- and membership of a chronological Group.
Sadeghi explains that placing Q 96:1–5 as the first Block is a historical exception to Bazargan’s statistical ordering. In his analysis the passage is then placed with the Blocks in the region of Group 3. Sadeghi, “The Chronology of the Qurān,” pp. 235–236.
8.7. Chronological relations hold on average
If Group 8 is placed after Group 7, this must not be read as a claim that every verse in Group 8 was certainly revealed after all the verses in Group 7. The more accurate formulation runs:
On average, the passages in Group 8 are placed later than the passages in Group 7 on the ordering under test.
The distinction is needed because:
- the date of each Block is not known with certainty;
- some Blocks may exchange positions without altering the general pattern;
- Group boundaries are not calendar boundaries;
- and stylometric testing works better on broad textual tendencies than on individual verses.
Sadeghi subsequently found stronger support for seven large clusters: Group 2, Group 3, Group 4, Group 5, the combination of Groups 6–11, the combination of Groups 12–19, and the combination of Groups 20–22.
This means that his stylometric data support the development between large phases more strongly than they confirm the exact order of each neighbouring Group. The internal chronology of the Groups within one large phase remains open to debate.
8.8. The place of Groups within the method as a whole
The position of chronological Groups within the method may be summarised as:
verses → Segments → Blocks → ṭūl-e mabnā → the order of 194 Blocks → 22 chronological Groups
The determination of years, meanwhile, follows a different path:
Block order → cumulative tokens → historical calibration curve → estimated year
The two paths are connected without being one and the same. A chronological Group is an aggregation of positions within an order, while a year of revelation is an estimated position on a scale of time. A Group is thus a way of dividing the order, and a year a way of dividing time.
8.9. Presentation in MushafNuzuli
To prevent Group numbers from being misread as year numbers, each Group page should display the following data separately:
| Data | Example presentation |
|---|---|
| Group number | Group 8 |
| Members | Blocks 111–116 |
| Relative position | Middle period |
| Estimated mission year | Around years X–Y |
| Historical period | Meccan or Medinan |
| Basis of dating | Bazargan’s calibration |
| Degree of certainty | An estimate, not a fixed date |
The recommended label reads “Group 8 — estimated to belong to the region of years X–Y of the prophetic mission”, and not “Group 8 — the eighth year of revelation”.
For a tooltip, the following brief definition may be used:
A chronological Group is a combination of several Blocks adjacent in Bazargan’s order, formed for the purposes of analysis. The Group number indicates relative position, not the year of revelation.
8.10. Conclusion
Chronological Groups form an analytical layer between individual Blocks and the broader phases of history. The 22 Groups were formed by Sadeghi by combining the 194 Blocks adjacent in Bazargan’s order.
Their correlation with the year of revelation is positive but not one-to-one. Higher-numbered Groups generally represent later material, while the Group number still cannot be read directly as a year number.
The most accurate formulation is:
A chronological Group is a cluster of adjacent Blocks that on average represents a consecutive region of time. Its relation to the year of revelation must be established through the historical calibration curve; a Group may therefore fall within one year, cross a year boundary, or span several neighbouring years.
In other words: a Group indicates relative position, while calibration indicates an estimated year.
For more accurate attribution, the full term may be written as Sadeghi’s 22 chronological Groups based on Bazargan’s ordering of the Blocks. The shorter label “Bazargan’s Groups” remains usable, with the caveat that the division into 22 Groups is Sadeghi’s analytical clustering of the chronology of 194 Blocks that Bazargan established.