Your thesis regulations almost certainly contain a phrase like presented in clear academic English. That phrase is not a criterion. Nobody can apply it identically twice, nobody can defend it at an appeal, and in practice each supervisor substitutes their own preferences for it — which is why two candidates in the same department receive contradictory instructions about the same paragraph and both believe they are following the rules.
A departmental writing standard is the document that replaces that phrase with something enforceable. It is a marking-consistency instrument, not an expression of taste, and if it is written as taste it will be ignored.
Ten steps, each with a named owner and the artefact it produces. Two afternoons of drafting and one committee cycle is a realistic timeline.
Step 1: Write the scope sentence before anything else
Owner: graduate school director or director of postgraduate research. Artefact: one sentence at the head of the document.
The sentence has to say what the standard is for and what it is not for, because every subsequent argument in committee will be settled by reference to it. Something in this shape:
This standard sets the written presentation requirements that examiners in this department will apply consistently to a thesis. It does not govern the quality of the research, and it does not create a separate ground for failure.
The second half is not decoration. Without it, someone will eventually try to use the standard as a route to fail a thesis whose research is sound, and you will lose that appeal.
Step 2: Name the style authority and pin the version
Owner: department, on advice from the subject librarian. Artefact: a clause naming one authority, one edition, and one fallback.
Naming a published style is nearly free. The open Citation Style Language repository carried 2,858 style files at its root when we read it on 23 August 2026, and it is the shared source that most reference managers draw on, so whichever mainstream style you name, the tooling your candidates already use can almost certainly produce it.
Naming a house variant is the expensive choice. A local modification means somebody has to author and maintain a style file, and it means every candidate has to install it. Do it only where a genuine disciplinary requirement makes a published style unusable, and if you do, name the person who owns the file.
Theses in this department are referenced in [style, edition]. Where this standard and the style manual conflict, this standard prevails. Candidates may use any reference manager that outputs the named style.
Step 3: Split the document into mandates and recommendations, and keep the mandate list short
Owner: the drafting group. Artefact: two clearly labelled lists.
The test for a mandate is brutal and worth applying literally: could two markers who have never spoken apply this clause to the same chapter and reach the same conclusion? If not, it is a recommendation.
| Usually a defensible mandate | Usually a recommendation |
|---|---|
| Reference style and edition | Paragraph length |
| Word limit and what it excludes | Use of the first person |
| Required front matter, in order | Sentence length |
| Language variety, consistently applied | Signposting phrasing |
| Figure and table numbering and captions | Choice of hedging expressions |
| File format and accessibility requirements at deposit | Verb tense within a section |
Note where verb tense and hedging fall. Both are real features of competent academic prose and neither survives the two-marker test, because the correct choice depends on what the sentence is doing. Put them in the recommendations with worked examples, where they will teach, rather than in the mandates, where they will be litigated.
Step 4: Give each chapter a structural expectation, in one table
Owner: two experienced examiners from the department. Artefact: a one-page table.
This is the part candidates actually use, and the part that reduces supervisor load most, because it answers in a table the questions that currently arrive one email at a time. For each chapter, state its function in one sentence, its indicative length as a percentage of the whole, and the single most common failure examiners see.
Keep it descriptive. The discussion interprets results against the reviewed literature and does not introduce new analyses is a structural expectation. The discussion should be engaging is not.
Departments that publish this table consistently report the same second-order effect: supervisors stop spending review time on structural questions the table has already answered, which is the workload pattern described in what your supervisors are actually reviewing.
Step 5: Write the standard-of-English clause with the equality office in the room
Owner: graduate school with the disability and equality service. Artefact: one clause plus a named referral route.
This is the highest-risk clause in the document. It has to do three things at once: state that a thesis must be intelligible to an examiner, keep language quality away from the misconduct route, and name where a candidate goes for help before it becomes a problem.
A thesis must be intelligible to an examiner in the field. Where language issues obscure the argument, the examiners may require corrections. Language quality alone is not a ground for a finding of academic misconduct, and reasonable adjustments apply in the usual way. Support is available through [named service] and candidates are encouraged to use it before submission rather than after.
Two failure modes to design against. The first is a standard that becomes a proxy for a nationality or disability screen; the exposures that creates are set out in why your international cohort needs the most writing support and is offered the least. The second is a standard that promises support the institution does not fund, which converts a policy document into a documented gap.
Step 6: Fix the language variety and say what consistency means
Owner: department. Artefact: one line.
Pick one variety of English and require internal consistency rather than a particular choice. A document that mixes varieties is the single most visible inconsistency in a thesis, because it appears on almost every page, and it is the cheapest thing on this list to check. One line does it:
Either British or American spelling and punctuation conventions are acceptable. One convention must be applied consistently throughout, including in figure captions, table headings and appendices.
Step 7: Point at the AI rules; do not restate them
Owner: whoever owns the institutional AI policy. Artefact: a cross-reference, not a copy.
The temptation is to write an AI section into the writing standard. Resist it. Two documents saying similar things about the same conduct will diverge at the first amendment, and a candidate will then be able to point at the more permissive one.
Write a cross-reference and nothing more. Note also that doctoral work and coursework need different rules for reasons that have nothing to do with writing standards, which is worked through in how AI rules should differ for a doctoral thesis compared with coursework.
Step 8: State what happens when the standard is not met — and what does not
Owner: the examinations officer. Artefact: a short enforcement clause.
Three routes, in escalating order, and a fourth that is explicitly excluded:
- Pre-submission. The supervisor flags it; the candidate fixes it. No record.
- At deposit. The formatting and metadata check returns it. This is a real cost and it lands on a conferral date rather than on staff hours, as set out in why your deposit desk is holding degrees for metadata.
- At examination. The examiners include presentation items in corrections.
- Excluded: a failure recommendation on presentation grounds alone, and any route from a presentation defect to the misconduct process.
Step 9: Version it, date it, and write the applicable-version rule
Owner: graduate school. Artefact: a version block and one sentence.
A thesis takes three to seven years. Your standard will change during it. Without a rule, a 2023 starter and a 2026 starter will each be told they are following the wrong document, and the appeal will be decided in the candidate’s favour.
Version [n], approved [date], effective [date]. Candidates are examined against the version of this standard in force at the date of their registration, unless they elect in writing to be examined against a later version.
The election clause matters: a later version is often more permissive, and a candidate should be able to take the benefit of it.
Step 10: Publish it where both the candidate and the marker will find it, then test it
Owner: graduate school administrator. Artefact: a published page and a test result.
Two tests before you call it done. First, hand the finished standard to two examiners with the same real chapter and ask each to list the presentation items they would raise. If the lists diverge substantially, your mandates are still recommendations wearing the wrong label. Second, send it to the deposit desk and ask whether a thesis that complies with it will pass their check — because if the standard and the deposit requirements disagree, candidates will discover it at the worst possible moment.
Then brief supervisors, because a standard that supervisors have not been walked through gets contradicted in the first meeting. The mechanics of running that briefing at scale are covered in how to train supervisors to use and govern writing tools.
What this costs to run once it exists
The document is cheap. The consistency is not. A standard raises the question of who checks compliance, and the honest answers are: the supervisor, at a cost in review time; the deposit desk, at a cost in conferral delay; or the platform the candidate writes in, at no marginal staff cost.
Tesify for Institutions holds a named reference style and a chapter structure inside the drafting environment, so compliance happens while the thesis is being written rather than being audited afterwards. That is the difference between a standard that changes documents and a standard that generates corrections.
If you are drafting or revising a writing standard this cycle, request an institutional evaluation and we will scope a free departmental pilot against the standard you are writing, with the compliance criteria agreed before it starts.
Frequently asked questions
What is a departmental writing standard?
A short published document that turns a phrase like clear academic English into presentation requirements examiners can apply consistently. It is a marking-consistency instrument rather than a statement of style preference, and it should say explicitly that it does not create a separate ground for failure.
Should the standard sit at department or institution level?
Both, in layers. The institution sets what is common across all theses, such as deposit format and the misconduct exclusion. The department sets what is disciplinary, such as the reference style and the chapter structure table. A single institutional document either becomes too vague to use or overrides legitimate disciplinary differences.
How long should it be?
Four to six pages, including the chapter table. Anything longer stops being read, and a standard nobody reads produces exactly the inconsistency it was written to remove.
Which reference style should we name?
Whichever one your discipline publishes in, named with its edition. The choice is nearly cost-free because most reference managers draw on the open Citation Style Language repository, which carried 2,858 style files at its root when we read it on 23 August 2026. A local house variant is the expensive option and needs a named owner.
Should we mandate verb tense by chapter?
No. Tense conventions are real and worth teaching, but the correct choice depends on what a sentence is doing, so two independent markers will not reach the same conclusion. Put tense in the recommendations with worked examples rather than in the mandates.
Can a thesis be failed for poor written English?
Your standard should say it cannot on that ground alone. Examiners may require corrections where language obscures the argument, but language quality must not be a route to a failure recommendation or to the misconduct process, and reasonable adjustments apply in the usual way.
How do we stop the standard becoming a screen on international candidates?
Draft the standard-of-English clause with the equality and disability service in the room, keep it to intelligibility rather than fluency, name a support route, and monitor corrections rates by cohort. If the clause produces corrections that concentrate on one cohort, it is doing something other than what it says.
Should the writing standard cover AI use?
Only by cross-reference. Two documents governing the same conduct will diverge at the first amendment, and a candidate will then be entitled to rely on the more permissive one. Point at the AI policy and stop there.
What happens to candidates who started before the standard existed?
State the rule explicitly: candidates are examined against the version in force at their registration date, unless they elect in writing to be examined against a later one. Without that sentence, the point is arguable and the appeal usually goes to the candidate.
Who owns the standard once it is approved?
Name a role rather than a person, normally the graduate school director, with an annual review date. An unowned policy document drifts out of alignment with the deposit requirements within two cycles.
How do we know the standard is actually working?
Give two examiners the same real chapter and compare the presentation items each would raise. Substantial divergence means your mandates are still recommendations in disguise. Repeat the test after each revision.
Does a writing standard reduce supervisor workload?
It reduces the share of review time spent on structural and formatting questions, because a published chapter table answers them once instead of one email at a time. It does not reduce the time spent on argument, which is the time you want supervisors to spend.