The Order of Writing

Twenty-seven papers, sequenced by what each one needs, what it risks, and what it earns you for the next.

Papers 27 Waves 4 Writable now 3 Realistic horizon 8–10 years for all

Twenty-seven papers is a menu, not a plan. Working alongside a business, without an institution or a research assistant, two to three papers a year is a good rate and four is a very good one. Twenty-seven is therefore roughly a decade — and nobody should commit to a decade of anything on the strength of one evening.

Choose six. The sequence below is ordered so that the first six are the ones worth choosing, and so that stopping after any wave still leaves you with a coherent body of work rather than a pile of fragments.

Three principles decide the order.

1. Narrow claims first. An undiscovered error damages a paper in proportion to what it claims. A dataset paper survives a correction as a version update; a paper claiming a finding does not survive it at all.

2. Build the platform before standing on it. You have no institutional affiliation and no publication record. Two citable outputs change how the third submission is read — not because the work is better, but because a reviewer has something to check you against.

3. Engineering before argument. Several papers need a month of building before a word can be written. Those months belong in the schedule, not in the surprise.

Wave 1 — The platform

months 0–6 · already written or nearly

Three papers that exist in draft today. None makes an interpretive claim, so none can be destroyed by an error in interpretation. Their job is to make you a citable author before you attempt anything contestable.

The Qurʾānic root concordance as an open dataset

topic 9

Venue: Journal of Open Humanities Data
Needs: Nothing. It is finished. Write the 2,000-word data descriptor and deposit the files
Claim: “This annotation contains these roots.” The narrowest claim you will ever publish
Earns: A DOI, a citable dataset, and the thing every later paper cites in its methods section

Counting words in the Qurʾān: root, lemma, sense

topic 20

Venue: Digital Scholarship in the Humanities
Needs: Two weeks. The worked examples and failure modes exist
Claim: A count without a stated level is meaningless. Survives even if a worked example turns out wrong
Earns: The methodological ground every counting paper afterwards stands on — including other people’s

Twenty-one defects: auditing in humanities computing

topic 21

Venue: Digital Scholarship in the Humanities
Needs: Three weeks. The register is written; the analysis of machine-versus-human detection is the contribution
Claim: A report of what went wrong and what caught it. Almost nobody publishes this, which is why it is worth publishing
Earns: A reputation for candour, which is the only currency an unaffiliated researcher actually has

Stop here and you have a real body of work: a citable dataset, a method others will use, and an honest account of your own failures. Many academics do not manage that in three years.

Wave 2 — The tafsīr corpus

months 6–20 · your rarest asset

Twenty-one aligned commentaries spanning a millennium. The digital-humanities literature names this work as an opportunity not yet taken. Each paper here needs engineering first, and the engineering compounds — paper 4 builds the tooling that papers 5 and 6 then use.

Exegetical attention: what commentators spend words on

topic 3

Venue: Journal of Qurʾanic Studies
Build first: Per-āyah length normalisation across 21 editions of very different scale. ~1 month
Why first in this wave: Purely descriptive, no interpretation, and it produces the normalisation every later tafsīr paper needs

Where the mufassirūn diverge

topic 1

Venue: Journal of Qurʾanic Studies; DSH
Build first: A divergence measure over commentary vocabulary. ~2 months, reusing paper 4’s normalisation
Discipline: Report where the tradition divides, never who is right. The moment you adjudicate, you need an ijāzah you do not have

The 399 hapax roots against gharīb al-Qurʾān

topic 11

Venue: Journal of Qurʾanic Studies; Arabica
Build first: Little. The hapax list exists. The work is reading the classical gharīb literature — months of it, and it cannot be automated
Why it matters: Puts a computational result into direct conversation with a classical genre. This is the paper that makes you legible to traditional scholars

Classical against modern exegesis, measured

topic 4

Venue: Journal of Islamic Studies
Build first: Nothing new — reuses papers 4 and 5
Risk: Editions differ in register and length; the normalisation must be argued, not assumed

Wave 3 — Comparison and structure

months 20–42 · needs the platform to exist

Now you have a record, a method others cite, and working tafsīr tooling. These papers make real claims, and they are safe to attempt only because a correction would no longer be your first public act.

Verse length and root density, replicated

topic 16

Venue: Arabica
Why here: Replication of Bazargan and Sadeghi on newer data with open code. Frame it honestly as replication — and it earns you the right to be read by the people who will review paper 11

Ḥadīth density per āyah in two traditions

topic 6

Venue: Journal of Shiʿa Islamic Studies
Build first: Āyah–ḥadīth linkage, route-and-confidence tagged. ~2 months
Discipline: Publish distributions. Do not characterise, rank or evaluate either tradition. The map is the finding

Shared vocabulary of the two ḥadīth corpora

topic 7

Venue: DSH; Journal of Shiʿa Islamic Studies
Build first: Reuses paper 9’s pipeline

The content-lexical signature of the revelation phases

topic 10

Venue: Arabica — where Sadeghi published
Needs: To have read Sadeghi 2011 properly, plus Bazargan and Bannister. Not cited — engaged
Why so late: This is the paper you wanted to write first. It is the most exposed thing on the list: an error in the measure destroys the finding outright. Attempt it once you have four papers behind you

Rhyme and verse endings, computationally

topic 15

Venue: Journal of Arabic Literature
Build first: Fāṣila extraction. ~1 month

Formulaic density, replicated on the morphological layer

topic 18

Venue: DSH
Prior work: Bannister. Cite throughout, replicate honestly

The sīrah corpus as a comparison set

topic 26

Venue: Journal of Islamic Studies
Build first: Little — the 17 works are cleaned already

Numerical claims: a reproducible audit

topic 27

Venue: JQS; DSH
Why not earlier: Mostly written, but it touches a popular literature and invites argument. Publish it when your record can absorb that

Semantic fields across the sūrahs

topic 12

Venue: JQS
Risk: Field assignment is interpretive. Declare it as data, itemised, the way the sense layer already is

Ring composition, tested against a null model

topic 17

Venue: Journal of Arabic Literature; JQS
Note: The null model is the paper. A rigorous negative result is a genuine contribution here

Wave 4 — Requires something you do not yet have

year 4 onward · collaborators, tooling, or training

Every paper here is blocked on an acquisition, not on time. Listed so the blockers are explicit rather than discovered late.

#	Paper	Topic	Blocked on
18	Ḥadīth quoted inside the tafsīrs	5	Reliable fuzzy matn matching at scale
19	Citation network of the mufassirūn	2	Arabic named-entity recognition and name normalisation
20	Sense-disambiguated lexicon, released	22	A year of annotation, ideally funded
21	Matn overlap between the two corpora	8	Same matching problem as 18, plus considerable delicacy
22	Qurʾān against pre-Islamic poetry	13	A comparable poetry corpus — OpenITI may serve
23	Language models on Classical Arabic morphology	23	Nothing technical — but the field moves fast. Publish quickly or not at all
24	The muqaṭṭaʿāt	19	A pre-registered protocol and the discipline to publish a negative result
25	Loanwords and foreign vocabulary	14	An angle Jeffery did not take. Distribution across phases, perhaps
26	Syriac and Biblical intertextuality	25	Syriac and Hebrew, or a collaborator who has them
27	Manuscript variants	24	A manuscript specialist. Do not attempt alone

If you only do six

Order	Paper	Why this one
1	Concordance dataset	Finished. Lowest risk. Everything else cites it
2	Counting: root, lemma, sense	Nearly written. Becomes other people’s method
3	Twenty-one defects	Nobody publishes this. That is exactly why it lands
4	Exegetical attention	Opens the tafsīr corpus, your rarest asset
5	Hapax against gharīb	Makes you legible to traditional scholarship
6	Where the mufassirūn diverge	The most interesting thing you can honestly claim

Three rules for the whole run

Never start a paper before the literature search. The Makkan and Madinan finding was called a thesis point when Sadeghi had published its substance in 2011. One evening of searching would have caught it. Do that evening first, every time.

Publish the code and data with every paper. It is how someone else finds the error you missed — which is a service to you, not a threat, provided it happens before a reviewer does it.

Keep the defect register open and public. Six of twenty-one defects were caught by a human reading, and no automated check could have caught any of them. The register is what turns that into a method rather than an anecdote.