---
name: examiner-proofing-and-defence-strategy
description: >
  Reads a thesis as a hostile expert would, finds the standard attack surfaces
  before the examiner does, and decides for each whether to fix it, own it in the
  text, or prepare a defence. Covers doctoral examination strategy, the viva
  itself and the corrections process. Use for "prepare for my PhD viva", "what
  will the examiner attack", "read my thesis critically", "defend my thesis",
  "how do I answer a question I cannot answer", "known examiner preparation",
  "pre-submission hardening", "major corrections response", "which weaknesses do
  I fix and which do I own".
category: 15 Academic University Research
ref: "15.19"
tier: 3
inherits: [K2, K3, K4, K5]
---

# Examiner-Proofing and Defence Strategy

## 1. One-line description

A method for reading your own thesis as a hostile expert, finding the specific points an examiner will attack, deciding for each whether it must be fixed before submission, owned explicitly in the text, or defended in the room, and preparing an examination performance in which conceding real points well is part of the defence.

## 2. What this skill is used for

**The research problem it solves.** Candidates prepare for examination by rehearsing what they want to say about their thesis. Examiners arrive having read it looking for what is wrong with it, which is their job. The gap between those two activities is where vivas go badly, and it is entirely closable, because the attack surfaces are not mysterious: the unjustified method choice, the construct that shifts definition, the claim that exceeds the data, the literature not read, the alternative explanation not addressed, the contribution smaller than claimed, and the limitation understated. Every one is findable in advance by a reader willing to be hostile to their own work, which is a learnable skill and the highest-return activity in the final months. This skill installs that reading, converts each finding into a decision (fix, own, or defend), and prepares the defence for what remains, on the principle that every thesis has weaknesses and a candidate who pretends otherwise has just demonstrated the largest one.

**Where it sits in the research lifecycle.** In the final months before submission, where findings can still be fixed; immediately before the examination, where the defence is rehearsed; and after, where corrections are responded to.

**Typical use cases.**
- A structured adversarial read of a complete draft, three to six months before submission.
- Deciding, weakness by weakness, what to fix and what to own.
- Preparing the contribution statement to withstand direct challenge.
- Anticipating specific examiners from their published positions.
- Rehearsing the viva, including the questions with no good answer.
- Preparing to concede a real point without conceding the thesis.
- Responding to a corrections list professionally, including where you disagree.

**Who uses it.** Doctoral candidates in the final phase; supervisors running mock examinations; researchers preparing for any expert defence of a substantial body of work.

## 3. When to use it

- You have a complete or near-complete draft and want its weaknesses found while there is still time to act.
- You are three to six months from submission and want to know what to spend that time on.
- Your examiners are known or likely and you need to prepare against their specific positions.
- You have a viva date and need structured preparation rather than re-reading your thesis.
- There is something in your thesis you are hoping nobody notices.
- You need to decide whether a weakness should be fixed, disclosed in the text, or defended verbally.
- You have received corrections and need to respond to them, including to points you think are wrong.
- You are supervising a candidate and running a mock defence.

## 4. When NOT to use it

- **Boundary with 15.12 Thesis Defence and Viva Preparation.** 15.12 prepares a masters-level defence: anticipating examiner questions on a dissertation, stress-testing weak points and preparing defensible answers, in a shorter examination against a standard of competent independent research. This skill takes over for doctoral examination, where the standard is an original contribution defensible against subject experts, the examination is longer and more adversarial, and, critically, the work includes pre-submission hardening of the thesis itself rather than preparation for a defence of a fixed document. In the other direction, this skill is disproportionate for a masters viva: running a full adversarial audit and examiner-position analysis on a dissertation examined for competence rather than originality wastes preparation time that should go elsewhere. The deciding question is whether the examination will test an original contribution against expert challenge.
- **The draft is too incomplete to attack.** An adversarial read needs the argument in its final shape. Below a substantially complete draft, the useful work is structural, and that is **15.18 Thesis Architecture and Chapter Coherence**.
- **The problems found are structural rather than defensive.** Where the audit finds a research question never answered or a construct that drifts, the fix belongs to 15.18, which handles it as architecture. This skill identifies it as an attack surface and hands it over; it does not do the restructuring.
- **The contribution has not been established.** You cannot defend a claim that has not been stated and sized. Go to **15.13 Doctoral Research Positioning and Originality** first, because the first substantive question in most doctoral examinations is about the contribution and an unprepared answer sets the tone for everything after.
- **The purpose is reassurance.** This method is designed to find problems, and it will find them in every thesis, including good ones. A candidate who needs confidence rather than an audit should not run it in the week before the viva. Timing matters: run it early enough that findings can be acted on.
- **The weakness is a research integrity issue rather than a limitation.** Fabricated data, undisclosed conflicts, plagiarism or an unapproved ethical breach are not attack surfaces to be managed; they are matters for the institution's research integrity process. `RESEARCHER DECISION REQUIRED`, and the correct route is disclosure, not defence preparation (K5 §2.4).
- **The examination format is unknown.** Doctoral examination differs enormously: open or closed, public defence or private viva, with or without a written report in advance, with an independent chair or without, with the supervisor present or not, and with different correction categories and different outcomes available. Preparing for the wrong format is worse than general preparation, so establish the format first and do not assume one.
- **Academic integrity.** Per §12.1: this skill does not write thesis text or viva answers for the candidate to deliver as their own unaided work. It generates the questions, locates the weaknesses and specifies what an adequate answer must contain; the candidate constructs the answers. Where an institution prohibits AI assistance for a task, this skill must not be used for it.

## 5. Required inputs

**Required.**
- **The thesis, complete or substantially complete.** An adversarial read cannot run on a summary or on the candidate's account of the work, because the gap between what a candidate believes their thesis says and what it says is precisely where the attacks live (K4 §6.4).
- **The contribution statement as it will be defended.** The examination's central question, and everything else is checked against it.
- **The examination format and the institution's outcome and correction categories.** These determine what a bad outcome actually is and what preparation is proportionate.

**Optional, and what each one adds.**
- **The examiners' names and their published work.** Converts generic defence into specific defence: what they have argued, what methods they favour, what they have criticised in print, and whether the thesis contradicts them. This is the single highest-value optional input.
- **The supervisor's own list of concerns.** Supervisors usually know the weak points and often have not said them plainly. Asking directly for the three things they would attack is one of the most useful conversations available.
- **Previous examiner reports or viva accounts from the department.** Reveal local norms: how long, how adversarial, which questions recur, and what the panel tends to care about.
- **The upgrade or confirmation report and its feedback.** Points already raised once will be raised again if not resolved, and an unresolved earlier criticism is an aggravating factor.
- **Any published papers from the thesis and their peer review.** Reviewer comments are a free preview of expert objections, and any that were not addressed will surface again.
- **A colleague willing to conduct a mock examination.** The performance element cannot be rehearsed alone.

## 6. Questions to ask before starting

1. **What is the one thing in this thesis you hope they do not ask about?** The fastest route to the primary attack surface, and candidates always know the answer. Default: if nothing comes to mind, the candidate has not yet read their own work adversarially, and Step 1 is where to start.
2. **What is the contribution, and what is the strongest argument that it is smaller than claimed?** Default: construct the deflationary case and test the answer against it.
3. **Who are the examiners and what do they hold?** Default: where unknown, prepare against the two or three most likely positions in the field, including the one most hostile to the approach taken.
4. **How much time remains before submission, and before the viva?** Determines whether findings become fixes or defences. Default: sort every finding by cost as well as severity.
5. **What does a bad outcome look like here?** Major corrections, resubmission and failure carry very different preparation implications, and the base rates differ by institution and discipline. Default: prepare to avoid resubmission, which is the realistic risk in most systems, rather than failure, which is rare.
6. **Which weaknesses are fixable and which are structural to the study?** A sampling limitation cannot be fixed after fieldwork; a missing justification can be written this week. Default: classify every finding on this axis before deciding anything.
7. **What will you concede, and what will you not?** Default: decide in advance, because a candidate who has not decided will either concede everything under pressure or defend everything, and both are damaging.

## 7. Step-by-step methodology

**Step 1. Read the thesis as an examiner who intends to find it wanting.**
This is a deliberate change of position and it does not happen by intending to be critical. Adopt a specific persona: a subject expert who did not want to examine this thesis, has read three others this year, is looking for the reason this one is not as good as it appears, and will be professionally embarrassed if they pass something weak. Read in the examiner's order, not yours: abstract, introduction, conclusion, methodology, findings, then literature. In that order the seams show, because the conclusion is read before the evidence and every claim in it must be recognised later or it fails. Mark four things and nothing else: every claim you would not accept without more, every choice you would question, every place you would ask how you know that, and every place you were not convinced. Do not fix anything while reading. A good thesis generates thirty to sixty marks, and a low count means the persona was not adopted. *Correct result: a raw list of marks with locations, made in one adversarial pass with no repairs attempted.*

**Step 2. Sort the marks against the seven standard attack surfaces.**
Almost every doctoral objection is one of these, and sorting them makes the pattern visible:

| Attack surface | The examiner's question | What makes it land |
|---|---|---|
| Unjustified method choice | Why did you not do it the other way? | No comparative defence; a rejection reason that is really about access |
| Shifting construct | You define this differently in chapters 3 and 7 | The definition broadens between measurement and discussion |
| Claim exceeding the data | Your design cannot support that | Causal language on associational data; population claims from one site |
| Literature not read | Are you familiar with the work on X? | A relevant body absent, especially in an adjacent field or the examiner's own |
| Alternative explanation | Could this not be explained by Y instead? | A rival account the thesis never names |
| Contribution smaller than claimed | What exactly is new here? | Novelty asserted rather than established against nearest neighbours |
| Understated limitation | You say this is a minor limitation. Is it? | A limitation whose consequence for the conclusions is not stated |

A mark that fits none of these is usually either a presentational point or a genuine strength you doubted unnecessarily. Count the marks per surface: a cluster in one is diagnostic and tells you where the preparation must concentrate. *Correct result: every mark classified, with the count per surface and the dominant cluster identified.*

**Step 3. Assess each finding on severity and on fixability, which are independent.**
Severity is what an examiner would do with it: could it produce a resubmission, would it generate a correction, would it produce a viva question you must answer well, or is it invisible to a competent reader? Fixability is what it would take to address: a paragraph, a section, a reanalysis, more data, or nothing because it is structural to a completed study. The two axes are independent and the combination determines the decision, which is why they must be assessed separately rather than merged into an impression. A high-severity, cheap-to-fix finding is the best thing this whole method produces. A high-severity, unfixable finding is what defence preparation exists for. *Correct result: every finding carrying a severity level and a fixability estimate, assessed separately.*

**Step 4. Decide fix, own, or defend for every finding, and be honest about the difference.**
Three decisions, each appropriate in different cases.

**Fix** where the weakness is real and removable in the time available. Missing justifications, unaddressed alternative explanations, over-claimed sentences, absent literature and unstated limitations are usually fixable and cheap. Over-claim in particular is often fixed by deleting three words, and it is among the most commonly attacked and most easily removed defects in any thesis.

**Own** where the weakness is real, cannot be removed, and would be worse hidden. Owning means stating it explicitly in the text, at the point where it matters, with its consequence for the conclusions and the conclusions adjusted accordingly. This converts an attack into a demonstration of judgement and is the single most under-used move available to candidates. An examiner who finds a limitation the thesis already states in the right place, with its effect carried through, has nothing to attack; one who finds a concealed limitation has found two problems.

**Defend** where the weakness is a reasonable disagreement, a defensible trade-off, or a limitation inherent to any study of this kind. Defending means preparing an argued answer, not deflecting: this was chosen over that, for these reasons, at this cost, and here is why the cost is acceptable given the question.

The failure mode is choosing defend for things that should be fixed because fixing is work. Test each defend decision by asking whether you would rather answer the question or not have to; if the honest answer is the latter and it is fixable, fix it. *Correct result: a decision per finding with a reason, and a scheduled fix list ordered by severity against cost.*

**Step 5. Harden the contribution statement against direct challenge.**
The contribution is the examination's central question, challenged in three predictable ways. That it has been done before, answered by the nearest-neighbour comparison from **15.13 Doctoral Research Positioning and Originality**, naming the closest studies and the specific difference. That it is smaller than claimed, answered by having already sized the claim to the evidence, so the deflationary version is your version. And that it is not a contribution at all but only competent work, answered by naming what is known now that was not known before, in one sentence, without describing the fieldwork. Rehearse at three lengths: thirty seconds, two minutes, and a full discussion. The thirty-second version matters most and is least often prepared, because it opens most doctoral examinations and sets the register for everything after. *Correct result: a three-length contribution answer, rehearsed aloud, with the deflationary challenge answered in each.*

**Step 6. Prepare against the specific examiners, where they are known.**
Read their recent relevant work and identify three things. What positions they hold that your thesis bears on, since contradicting an examiner's published argument is entirely acceptable and requires that you have engaged their work at its strongest, cited it accurately, and can concede its merits before disagreeing. What methods they favour, since a methodologist examines methods and a theorist examines theory, and preparation weight should follow. And what they have criticised in others, a reliable predictor of what they will look for. Then check their relevant work is cited and cited correctly, since a missing or misrepresented citation to an examiner's own work is an avoidable and memorable failure. Two cautions. Do not over-fit: examiners ask about the thesis, not only about themselves, and a candidate prepared only for one examiner's hobbyhorse is exposed on everything else. And do not let anticipation become flattery, since transparent positioning to please an examiner reads badly and invites the challenge it was meant to avoid. *Correct result: a per-examiner note of positions, methodological preferences and likely questions, with citation accuracy verified.*

**Step 7. Build the question bank and write answers that concede where conceding is right.**
Generate questions in five bands. Opening: summarise the thesis, state the contribution, say why this topic. Methodological: every choice with an alternative, the sampling, the instrument, the analysis, the validity threats. Theoretical: construct definitions, the mechanism, the boundary conditions, why this framework. Findings and interpretation: the alternative explanation, the inconvenient result, the null, the base sizes, the generalisability. And the closing band, which candidates neglect: what would you do differently, what are the limitations, what next, what has this contributed. Write an answer to each, then apply three tests. Does it answer the question actually asked rather than an adjacent one you prefer? Is it under ninety seconds spoken, since long answers signal uncertainty and invite interruption? And does it concede what should be conceded? That last test is the difference between good and poor defence. Conceding well means accepting the point specifically, stating its actual consequence accurately, and stating what stands regardless. Conceding badly means denying an obvious problem, which destroys credibility for everything else, or over-conceding, which agrees the thesis is worthless because one criticism was correct. *Correct result: forty to sixty questions with answers under ninety seconds, each tested for directness, length and appropriate concession.*

**Step 8. Prepare the answers to questions you cannot answer.**
There will be some, and how they are handled matters more than most rehearsed answers. Three cases with different correct responses. A question about literature you have not read: acknowledge it directly, say what you do know of that area, and how you would expect it to bear on your work. Do not bluff, because an examiner asking about a specific literature usually knows it well and fabricated familiarity is discovered in one follow-up. A question about an analysis you did not run: say why, whether it would change the interpretation, and whether it is a limitation or a choice. A question exposing a genuine flaw: acknowledge it, state its actual scope precisely rather than minimising it, and say what remains true despite it. The general rule is that "I don't know" followed by how you would find out is a competent answer and far stronger than a confident wrong one. Thinking time is permitted and using it reads as considered rather than slow. *Correct result: a rehearsed protocol for each of the three cases, practised until it does not feel like failure.*

**Step 9. Run a mock examination under realistic conditions.**
Solitary preparation cannot rehearse performance. Use someone who has read the thesis, is willing to be genuinely adversarial, and preferably is not the supervisor, who is invested in the work and tends to defend it. Run it for a realistic duration and format, and record it. Then review for four things: answers that wandered, answers that were defensive where a concession was correct, answers over ninety seconds, and moments where you contradicted the thesis, which is the most dangerous and hardest to notice in the moment. A single mock finds performance problems no written preparation reaches, particularly the habit of answering the question you prepared for rather than the one asked. *Correct result: a recorded mock, reviewed against the four criteria, with the weak answers rewritten and re-rehearsed.*

**Step 10. Prepare the physical and procedural ground.**
Know the format, expected duration, who attends, the supervisor's role if present, whether you present at the start and for how long, what you may bring, and what outcomes and correction categories exist. Prepare a marked copy with tabs on the sections you expect to be taken to, since turning to a table in five seconds visibly steadies an examination. Prepare any expected opening statement, timed and rehearsed, because an opening that overruns starts badly. And prepare for the practical: examinations are long, concentration degrades, and asking for a question to be repeated or for a moment to think is normal rather than a weakness. *Correct result: format confirmed, thesis tabbed, opening timed, logistics settled in advance rather than on the day.*

**Step 11. Respond to corrections professionally, including where you disagree.**
Most doctoral examinations produce corrections, and they are a normal outcome rather than a judgement of failure. Build a response document with one row per correction: the correction as written, what was done, and where in the revised thesis it now appears with a page reference. Address every point, including those you consider minor, because an unaddressed correction is what triggers a second round. Disagreement is legitimate and must be argued rather than ignored: state the point, give the reason, cite the evidence, and where possible offer a partial accommodation such as adding the examiner's preferred interpretation alongside your own rather than replacing it. Never simply skip a correction and hope it is not checked, because it will be, and an examiner who finds an ignored point reads it as a candidate who did not engage. Keep the tone neutral: these documents are read by people who spent hours on your thesis, and defensiveness is visible in prose. *Correct result: a complete correction-by-correction response table with page references, and every disagreement argued rather than passed over.*

## 8. Analytical framework

Every weakness passes through the same decision structure:

    Finding → Attack surface it belongs to → Severity (what an examiner would do) →
    Fixability (what it would cost) → Decision (fix, own, defend) →
    Action or prepared answer

The two assessments must stay separate. Merging severity and fixability into a single impression produces the two characteristic errors: fixing cheap trivia because it is cheap while leaving an expensive severe finding untouched, and deciding to defend something that could have been fixed in an afternoon.

The defence itself has its own structure, and answers that omit a step are the ones that fail:

    Acknowledge the point → State its actual scope →
    State what stands regardless → Offer what would resolve it

Omitting the first step reads as defensiveness. Omitting the second either minimises a real problem or inflates a small one. Omitting the third concedes the thesis along with the point, which is the commonest way a candidate loses ground they did not need to lose. Omitting the fourth is a missed opportunity, since naming the study that would settle the question turns a weakness into a research agenda.

## 9. Output format

**1. Adversarial read log.** Every mark with its location, made in one pass.

**2. Attack surface analysis.**

| Finding | Location | Attack surface | Examiner's likely wording | Severity | Fixability | Decision |
|---|---|---|---|---|---|---|

**3. Fix schedule.** Ordered by severity against cost, with dates, and a cut line marking what will not be done.

**4. Own list.** Each weakness to be stated in the text: where it goes, how it is worded, and which conclusions are adjusted as a result.

**5. Contribution defence.** The three-length answer, with the three standard challenges and their responses.

**6. Examiner analysis** where examiners are known: positions, methodological preferences, likely questions, citation accuracy check.

**7. Question bank.** Forty to sixty questions across the five bands with answers under ninety seconds, marked for where a concession is required.

**8. Unanswerable question protocol.** The three cases with rehearsed responses.

**9. Mock examination review.** Recording notes against the four criteria, with rewritten answers.

**10. Logistics checklist.** Format, duration, attendees, opening statement, tabbed copy, outcome categories.

**11. Corrections response template** for post-examination use.

**When the evidence is thin, the format must not force fabrication (K4 §1).** Where examiners are unknown, the examiner analysis is omitted rather than populated with guesses about people. Where an adversarial read finds few problems, that is reported honestly, though it far more often means the persona was not adopted than that the thesis is unusually strong, and re-reading with a sharper persona is the correct response. Where a weakness has no good defence, the output says so and moves it to the fix or own list rather than manufacturing a defence that will not survive a follow-up question.

## 10. Quality checks

Run before submission and again before the examination. These sit on top of K4 §8.

1. Was the adversarial read done in the examiner's reading order, in one pass, with no repairs attempted?
2. Does the read log contain enough findings to be credible, or was the persona not adopted?
3. Is every finding classified against an attack surface, and is the dominant cluster identified?
4. Are severity and fixability assessed separately for every finding?
5. Is every "defend" decision genuinely a defensible trade-off rather than an avoidance of work?
6. Has every over-claimed sentence been located and cut, particularly causal language on non-causal designs?
7. Does every limitation now state its consequence for a named conclusion rather than merely existing?
8. Can the contribution be stated in thirty seconds without describing the fieldwork?
9. Is the deflationary version of the contribution already the version in the thesis?
10. Where examiners are known, is their relevant work cited and cited accurately?
11. Does every prepared answer answer the question asked rather than an adjacent one?
12. Does every answer that should concede, concede specifically, and does it also state what stands?
13. Is there a rehearsed response for the case where you do not know?
14. Has a mock examination been run with someone other than the supervisor, and reviewed?
15. Is any finding being left unaddressed in the hope that it will not be noticed?

## 11. Common failure modes

| Failure | How to recognise it | How to prevent it |
|---|---|---|
| **Reading as an author** | The adversarial pass produces six mild findings | Adopt an explicit hostile persona (Step 1) |
| **Fixing while reading** | The read stops at the first problem | One pass, marks only, no repairs (Step 1) |
| **Defend as avoidance** | Expensive fixes reclassified as defensible choices | Test each defend decision against fixability (Step 4) |
| **Hiding a known weakness** | Something the candidate hopes is not noticed | Own it in the text at the point it bites (Step 4) |
| **Unsized contribution** | The contribution answer runs three minutes and describes methods | Rehearse the thirty-second version (Step 5) |
| **Over-fitted examiner preparation** | Only the examiner's own topic is prepared | Prepare the thesis, then the examiner (Step 6) |
| **Missing the examiner's own work** | An examiner's relevant paper uncited | Verify citation and accuracy (Step 6) |
| **Answering the adjacent question** | The answer is fluent and not responsive | Test every answer for directness (Step 7) |
| **The long answer** | Answers running past two minutes | Cap at ninety seconds and rehearse aloud (Step 7) |
| **Conceding nothing** | Every criticism is rebutted | Decide concessions in advance (Step 7) |
| **Conceding everything** | One correct criticism collapses the whole defence | Always state what stands regardless (Step 8) |
| **Bluffing on unread literature** | Vague familiarity offered under questioning | Use the acknowledge-and-relate protocol (Step 8) |
| **Contradicting the thesis in the room** | A verbal answer inconsistent with what is written | Record a mock and check for it (Step 9) |
| **Corrections ignored** | A correction not done and not explained | Address every point, argue disagreements (Step 11) |
| **Defensive corrections prose** | A response document that argues with the examiner's tone | Keep it neutral and evidential (Step 11) |

## 12. AI guardrails

Skill-specific. The universal prohibitions in K4 apply in full and are not repeated.

1. **Academic integrity, and it binds this whole skill.** These skills assist a researcher's thinking, structure and rigour. They do not produce work to be submitted as the student's own unaided output. The user must comply with their institution's AI use policy and its declaration requirements, which vary by institution and by assessment. Where an institution prohibits AI assistance for a task, this skill must not be used for it. The skill never writes a passage for submission as though the student wrote it; it interrogates, structures, critiques and teaches. Operationally here, the constraint is close to absolute: an examination is an oral test of whether the candidate's understanding is their own. The skill produces questions, locates weaknesses, and specifies what an adequate answer must contain and concede. It does not produce answers to be memorised, and a candidate reciting prepared text is visible within two follow-up questions.

2. **Never soften a finding to spare the candidate.** The value of this method is entirely in its hostility. A finding reported gently is a finding the examiner will report bluntly. Severity is stated as an examiner would act on it.

3. **Never fabricate an examiner's position, preferences or published views.** Where examiners are named, their positions come from work that has been read. Model knowledge about an academic's views is a standard source of confident error, and preparing against a position an examiner does not hold wastes the preparation and can produce an answer that insults them (K4 §2.4).

4. **Never manufacture a defence for an indefensible finding.** Where a weakness has no good answer, say so and route it to fix or own. A prepared defence that collapses under one follow-up question is worse than a prepared concession, because it costs credibility on everything else.

5. **Never advise concealment.** Not disclosing a known limitation, describing an exploratory analysis as confirmatory, or omitting an inconvenient result are integrity failures, not defence strategies (K4 §4.1, §4.3). The correct move for an inconvenient finding is always to own it.

6. **Never predict the outcome of an examination.** Outcomes depend on examiners, institution, discipline and the day. The skill assesses attack surfaces; it does not forecast results, and a reassuring prediction is both unfounded and actively harmful to preparation.

7. **Never write viva answers as deliverable scripts.** Specify what an answer must contain, what it must concede and what it must not give away, and let the candidate build it in their own words.

8. **Never treat conceding as losing.** Where a criticism is correct, the guidance is always to concede it accurately and state what stands. Advice to hold a position that the evidence does not support is advice to be caught doing so.

9. **Where a finding suggests a research integrity problem rather than a methodological weakness, stop and escalate** (K5 §2.4). This is not an attack surface to be managed.

## 13. Best-practice principles

- **Hostility to your own work is a skill, and it is the highest-return activity in the final months.** Every problem you find is one an examiner will not find first, and finding it early is what converts a resubmission risk into a paragraph of revision.
- **Read in the examiner's order, not yours.** Abstract, introduction, conclusion, methodology, findings, literature. The conclusion read before the evidence exposes every claim you cannot support.
- **Over-claim is the cheapest defect to fix and among the most attacked.** Most instances are removed by deleting a causal verb or a population noun, and a single pass hunting for them is worth the hour.
- **Owning a limitation in the right place removes it as an attack.** An examiner who finds a weakness already stated, with its consequence carried through into the conclusions, has nothing left to do with it.
- **Every thesis has weaknesses, and claiming otherwise is the largest one.** A candidate who cannot name their study's limitations has demonstrated that they cannot evaluate their own work, which is the capability being examined.
- **Conceding a real point well is part of defending well.** Accept it specifically, state its actual scope, say what stands regardless. Denial and collapse are the two failure modes and both are avoidable.
- **The thirty-second contribution answer is the most important sentence you will say.** It usually opens the examination and it sets the register for everything after.
- **"I don't know, and here is how I would find out" is a competent answer.** Bluffing is discovered in one follow-up question and costs credibility on everything already said.
- **Prepare the thesis first, then the examiners.** Examiner-specific preparation is valuable and must not displace preparation on the work itself.
- **A viva is a professional conversation, not an interrogation, and the examiners are usually interested.** Candidates who treat it as combat answer defensively and concede badly.
- **Corrections are normal.** They are the mechanism by which a thesis is improved, not a verdict on the candidate, and responding to them thoroughly and neutrally is straightforward professional work.
- **Disagree with a correction when you have grounds, and argue it.** Silent non-compliance triggers a second round; a reasoned response, with an offered accommodation, usually does not.

## 14. Worked example

Generic fictional scenario, academic, computing education.

**INPUT**

A doctoral candidate has a complete thesis on how novice programmers develop debugging strategies, based on a think-aloud study of 22 students and a follow-up survey of 180. Submission is in five months. Examiners are not yet appointed but the likely external is known to work on programming pedagogy and to be sceptical of think-aloud methods.

**PROCESS**

*Step 1.* The adversarial read is run in the examiner's order over two days. It produces 41 marks. Reading the conclusion before the findings immediately exposes the first: the conclusion states that novices "develop" strategies through a described sequence, but the study is cross-sectional and observed different students at different stages, which cannot establish development within individuals.

*Step 2.* Sorting shows a cluster of 14 marks under "claim exceeding the data", almost all of them developmental language on cross-sectional evidence. A second cluster of 7 sits under "unjustified method choice", all concerning the think-aloud protocol. The remainder are spread.

*Step 3 and 4.* The developmental language is high severity and cheap to fix: 14 sentences, mostly one verb each, replaced with language about differences between students at different stages. This is done in one afternoon and removes what would have been the examination's central attack. The think-aloud cluster is different. The method choice is defensible but the thesis never defends it comparatively, and it never addresses the standard objection that verbalising changes the cognitive process being observed. That objection is real, it cannot be removed after the fact, and the likely external is known to hold it.

*The judgement call.* Three options for the think-aloud objection: defend it verbally only, own it in the text, or run a small additional study comparing think-aloud with silent performance, which would take about six weeks of the five months available. The temptation is the third, because it would settle the point. The decision taken is the second, with preparation for the first. A new methodology subsection states the reactivity objection at its strongest, cites the literature on both sides including the work the external has published, explains what was done to mitigate it (practice tasks, minimal prompting, no explanation requests), and states plainly what it means for the findings: that the strategies observed are those available to verbalisation and that non-verbalisable processes are outside the study's reach. Two conclusions are then narrowed to match. The additional study is rejected because it would consume the time needed to fix the other 27 findings, and because a small underpowered methods study would create a new attack surface rather than closing one.

*Step 6.* The external's three most recent relevant papers are read. One argues a position the thesis's findings complicate. That paper is currently cited once in passing. It is now engaged properly across two paragraphs, its argument presented at its strongest, its merits conceded, and the specific point of difference stated with the evidence. This converts a probable hostile question into a prepared exchange.

*Steps 7 and 8.* A question bank of 52 questions is built. Two have no good answer: the study's single institution, and the absence of any longitudinal element. Both are moved to the own list and stated in the limitations with their consequences. The prepared answers concede both directly and state what stands: the strategy differences observed are real and were consistent across the survey sample, whatever their developmental trajectory.

*Step 9.* A mock with a colleague from a neighbouring department finds two problems no written preparation had: the contribution answer ran four minutes, and under pressure the candidate twice described the study as showing how strategies develop, contradicting the corrected text. Both are drilled.

**OUTPUT**

A thesis with 31 findings fixed, 4 owned in the text with conclusions adjusted, and 6 prepared as defences. A three-length contribution answer. A question bank with rehearsed concessions on the two genuinely weak points. And an engagement with the likely external's published position that is honest, accurate and complimentary before it disagrees.

`RESEARCHER SIGN-OFF REQUIRED` from the supervisor on the decision not to run the additional methods study, since it trades a stronger methodological position against the time needed for the remaining fixes (K5 §2.7).

## 15. Advanced usage

**The three-reader technique.** Run the adversarial read three times with different personas: the methodologist, who attacks design, sampling, measurement and analysis; the theorist, who attacks constructs, mechanisms and the theoretical claim; and the generalist second examiner, who attacks followability and asks whether the argument can be understood without specialist background. Each persona finds different problems, and the generalist read is the one most often skipped and most often needed, since a second examiner outside the sub-field frequently asks the hardest questions precisely because they will not accept what specialists take for granted.

**Preparing against a supportive examiner.** An examiner who likes the work is a different risk. They may accept claims that should have been challenged, leaving weaknesses undiscovered until publication or, worse, leaving the other examiner to do all the challenging. Prepare the same way, and be willing to volunteer a limitation that has not been raised, which reads as strength.

**Publication-based thesis examination.** The examination focuses disproportionately on the linking material and on the candidate's own contribution to co-authored papers, and both are tested directly. Prepare a specific, evidenced account of what you did on each paper, and be ready to defend the overarching contribution as distinct from the sum of the papers. See **15.17 Longitudinal and Multi-Study Design** and **15.18 Thesis Architecture and Chapter Coherence**.

**When a weakness is found too late to fix.** In the final weeks, a newly discovered structural weakness should be owned verbally rather than patched, since a rushed fix commonly breaks a cross-reference, contradicts another chapter or introduces an inconsistency that is more damaging than the original problem. Prepare to raise it yourself if the examiners do not, which almost always reads better than being shown it.

**Running a mock examination as a supervisor.** Be genuinely adversarial, which most supervisors find difficult on work they have shaped. Use the seven attack surfaces as a checklist, ask at least three questions the candidate cannot answer, and review the recording with them rather than only debriefing, because the performance problems are visible on a recording and invisible in memory. See **15.20 Supervision Feedback and Development**.

## 16. Skill chain

**Recommended previous skills:**
- **15.18 Thesis Architecture and Chapter Coherence.** Hands over the structural findings, which become attack surfaces here.
- **15.13 Doctoral Research Positioning and Originality.** Hands over the sized contribution and the nearest-neighbour comparison that defends it.
- **15.16 Doctoral Methodology Justification and Rigour.** Hands over the comparative defence of method choices and the validity threat register.
- **15.12 Thesis Defence and Viva Preparation.** The masters-level equivalent, for candidates whose examination tests competence rather than original contribution.

**Recommended next skills:**
- **15.24 Journal Article Development.** Takes the hardened contribution into publication, where the same objections arrive from reviewers.
- **15.26 Peer Review Response.** Uses the same concede-and-argue structure the corrections response uses here.

**Runs well alongside:**
- **15.20 Supervision Feedback and Development**, where a supervisor runs the adversarial read or the mock.
- **13.01 Research Quality Review** and **13.04 Bias Detection**, which audit the study independently of the candidate's own reading.
- **15.03 Referencing and Citation Management**, which verifies that examiners' own work is cited accurately.

---
A Yazi Supplied Skill and resource.