---
name: doctoral-methodology-justification-and-rigour
description: >
  Defends every methodological choice as a comparative argument against the
  alternatives that were rejected, checks that the philosophical position and the
  methods actually cohere, names validity threats specifically and addresses
  them, and installs the substantive practices behind qualitative rigour rather
  than the vocabulary. Use for "justify my methodology", "defend my method
  choice", "why this method and not that one", "validity threats", "epistemology
  and methods alignment", "trustworthiness criteria", "reflexivity statement",
  "theoretical sampling saturation", "pilot study changes", "pre-registration",
  "my methodology chapter is descriptive".
category: 15 Academic University Research
ref: "15.16"
tier: 3
inherits: [K2, K3, K4, K5]
---

# Doctoral Methodology Justification and Rigour

## 1. One-line description

A method for building a methodology argument at doctoral standard, where every choice is defended comparatively against the alternatives that were rejected, the philosophical position and the methods are shown to cohere, validity threats are named specifically and addressed rather than listed, and rigour is demonstrated through practices rather than asserted through vocabulary.

## 2. What this skill is used for

**The research problem it solves.** The difference between a masters methodology chapter and a doctoral one is not length or technical sophistication. It is that a masters chapter describes what was done and a doctoral chapter argues why this and not that. A description says "a cross-sectional survey was used". An argument says why a cross-sectional survey rather than a panel, an experiment, a case series or a document study, what each rejected option would have bought, what the chosen one costs, and what that cost does to the conclusions. Alongside this sits a second failure that examiners attack reliably: a philosophical position declared in chapter three and contradicted by the methods in the same chapter, usually because the position was chosen from a textbook list rather than derived from what the research question requires. And a third: qualitative rigour claimed in the vocabulary of trustworthiness with none of the practices behind it visible anywhere. This skill produces the comparative argument, tests the coherence, and demands the evidence of rigour rather than its language.

**Where it sits in the research lifecycle.** At design, where the choices are actually made and where the reasoning must be recorded before outcomes are known; and at write-up, where that reasoning becomes the methodology chapter. Also at examination preparation, where every unjustified choice is an attack surface.

**Typical use cases.**
- Converting a descriptive methodology chapter into a defended argument.
- Testing whether a declared paradigm and the chosen methods are actually compatible.
- Naming the specific validity threats a design faces and what was done about each.
- Building the audit trail, reflexivity and negative case work that qualitative rigour claims rest on.
- Defending a sampling decision, including the stopping rule for theoretical sampling.
- Documenting what a pilot changed, which is the evidence that piloting was real.
- Deciding whether to pre-register, and writing a protocol that constrains later choices.

**Who uses it.** Doctoral candidates in any discipline and any paradigm, writing or defending a methodology chapter; supervisors reviewing one before it meets an examiner; researchers preparing methods sections for peer review.

## 3. When to use it

- Your methodology chapter reads as a description of procedures rather than an argument for them.
- A supervisor or reviewer has asked "why did you not use X instead?" and you did not have an answer ready.
- You declared a philosophical position early and are no longer sure your methods are consistent with it.
- You have listed limitations and threats without saying what was done about them.
- You use the vocabulary of qualitative rigour and cannot point to where in the work each criterion was met.
- You need a defensible stopping rule for sampling, particularly theoretical sampling.
- You ran a pilot and are not sure what to report from it.
- Your field is moving toward pre-registration or data and code sharing and you need to decide what to do.
- You are preparing for examination and want the unjustified choices found before an examiner finds them.

## 4. When NOT to use it

- **The work is at masters level.** A masters methodology chapter is assessed on competent, transparent description with the main choices justified; running a full comparative defence of every decision produces a chapter out of proportion to the degree. Use **15.08 Research Design and Methodology Chapter**, which sets the standard a masters actually requires. In the other direction, this skill takes over where the chapter will be examined by subject experts who will ask what was rejected and why.
- **The design has not been chosen yet.** This skill defends a design; it does not select one. If the question is which method to use, that is **01.04 Research Method Selection**, and the output of that becomes the input to this.
- **The methodological problem is really a question problem.** A design cannot be justified against a research question that is vague, unanswerable or misaligned with the contribution. Where the justification keeps collapsing, the fault is usually upstream: return to **15.13 Doctoral Research Positioning and Originality** or **01.02 Business Problem to Research Question**.
- **The statistical analysis plan is what is needed.** Choosing and justifying tests, models, assumptions and handling of missing data is analytical planning. See **05.01 Statistical Analysis Planning**. This skill covers why the design generates data those tests can be run on, not which test.
- **Ethical approval is the task.** Risk assessment, consent materials, data management and vulnerable participant handling are **15.09 Ethics Clearance Application**. Ethics decisions constrain method choice and are referenced here, but the submission is a different document with a different audience.
- **The design has a fatal flaw that justification cannot repair.** Where the sampling frame cannot reach the population, where the instrument does not measure the construct, or where the design cannot answer the question, the correct output is that finding, not a better-argued defence of an inadequate design. A justification chapter that argues persuasively for a design that cannot work makes the examination worse, not better. `RESEARCHER DECISION REQUIRED` (K5 §2.7).
- **You want the paradigm chosen for you.** Philosophical positioning is a commitment with consequences throughout the thesis, and adopting one because it is common in the department or because it sounds sophisticated is exactly the failure that produces the incoherence attack. If the position cannot be argued from the question and the researcher's own commitments, the honest output says so.
- **Academic integrity.** Per §12.1: this skill does not write a methodology chapter for submission as the candidate's own unaided output. It interrogates choices, surfaces the alternatives that were not considered, and returns what a rigorous defence would require. Where an institution prohibits AI assistance for a task, this skill must not be used for it.

## 5. Required inputs

**Required.**
- **The research questions in their final form.** Every methodological justification is relative to a question. Justifying a design against a question that later changes produces a chapter that defends the wrong thing, which is a common and avoidable examination failure.
- **The design as it stands, including what was actually done.** Including the departures from plan, the access that failed, and the compromises made, because those are what the defence has to accommodate. A defence written against the idealised design is worthless.
- **The candidate's own statement of their philosophical position, in their own words.** Not a textbook paragraph. The position has to be one the candidate holds and can explain, because the viva will test whether it is understood or recited.

**Optional, and what each one adds.**
- **The alternatives that were considered and rejected, with dates.** Turns a retrospective rationalisation into a documented decision, which reads entirely differently to an examiner.
- **Pilot study materials and results.** Lets the chapter show what piloting changed, which is the only evidence that piloting was more than a formality.
- **The instrument and any validation evidence.** Determines whether validity can be argued from evidence or must be argued from construction.
- **Field notes, memos, coding records and the reflexive journal.** These are the audit trail. Without them, qualitative rigour claims are assertions.
- **The disciplinary and institutional expectations on transparency**, including whether data and code sharing or pre-registration are expected. These vary sharply and an assumption in either direction can be wrong.
- **Supervisor feedback and any panel comments on the methods.** Names the doubts that already exist, which is where an examiner will start.

## 6. Questions to ask before starting

1. **What did each research question require of the design, specifically?** Causal inference, prevalence estimation, mechanism identification, meaning-making, process over time. The requirement is the criterion against which alternatives are compared. Default: derive it from each question separately, because a design that suits question one often does not suit question three.
2. **What were the real alternatives, and why were they rejected?** Not straw alternatives. Default: reconstruct the plausible options honestly, including the one an examiner would have chosen, and answer it.
3. **What does the chosen design cost, and what does that cost do to the conclusions?** Every design forgoes something. Default: name the cost explicitly and carry it into the limitations and the claims (K3 §3.6).
4. **Is the philosophical position derived from the question, or adopted and then decorated?** Default: test coherence by checking whether the methods would still make sense under a different position, which usually reveals that the position is doing no work.
5. **Which specific threats does this design face, given this population and this measurement?** Generic threat lists are worthless. Default: instantiate every threat against the actual study or drop it.
6. **What is the sampling logic and what is its stopping rule?** Default: require the rule to be stated in advance and the evidence that it was met to be reportable.
7. **What does the discipline expect on transparency right now?** Pre-registration, data availability, code availability, reporting guidelines. Default: state the expectation assumed and flag it as needing confirmation, since expectations differ by field and are changing.

## 7. Step-by-step methodology

**Step 1. Derive design requirements from each research question separately.**
For each question, write what a design must be capable of in order to answer it: the type of claim (descriptive, associational, causal, interpretive, evaluative), the unit and level, whether temporal ordering is needed, whether the mechanism must be observed or only inferred, the precision required, and the population the claim must reach. This is the specification against which every option will be scored, and doing it question by question exposes the common problem of a single design bolted onto a set of questions with different requirements. Where the requirements conflict, that is a finding: it means either the questions need reducing or the study needs more than one component, which is the entry to **15.17 Longitudinal and Multi-Study Design**. *Correct result: a requirements table, one row per research question, with any conflicts between rows made explicit.*

**Step 2. Build the comparative justification, which is what makes it doctoral.**
For each major choice, list the genuine alternatives and evaluate each against the Step 1 requirements. Major choices include the overall design, the data collection method, the sampling approach, the instrument, the analytic approach and the timeframe. For each alternative, write what it would have delivered that the chosen option does not, what it would have cost, and why it was rejected: on validity, feasibility, ethics, access or paradigm fit. Naming which kind of reason it was matters, because a feasibility rejection is honest and a rejection dressed as principle when it was really about access is the kind of thing an examiner exposes. Include the option an expert examiner would most likely have chosen, and answer it directly. The comparative table is the single artefact that most reliably distinguishes a doctoral methodology chapter from a masters one. *Correct result: a comparison table per major choice, with rejection reasons categorised, including the examiner's likely alternative.*

**Step 3. Test philosophical coherence in both directions.**
State the ontological and epistemological position in the candidate's own words, then check it against the methods in two passes. Forwards: does this position imply these methods, or would it equally imply several others, in which case it is decorative? Backwards, the harder pass: do the methods as actually practised imply a different position from the one declared? The standard incoherences are recognisable. A constructivist position followed by an analysis that treats codes as counts and reports frequencies as prevalence. A critical realist position followed by an analysis that never moves past the empirical level to propose generative mechanisms. A post-positivist framing followed by an interpretive analysis of meaning with no measurement claim. A pragmatist position invoked to avoid the question rather than as a considered commitment to what works for the problem, which is what pragmatism actually claims. None is automatically fatal, and some are defensible if the tension is acknowledged and argued, but an unacknowledged one is among the easiest attacks an examiner has. *Correct result: a stated position, a forward and backward coherence check, and either alignment or an explicit argued account of the tension.*

**Step 4. Name validity threats specifically, and say what was done about each.**
Take the general threat categories relevant to the design and instantiate every one against this study, or drop it. For quantitative designs: which specific confound, and how it was addressed by design, measurement or analysis; what selection process actually operated in this recruitment; what the instrumentation risk is with this measure in this population; whether the attrition pattern is related to the outcome; what alternative explanation a critic would offer for the main result. For interpretive designs: where the researcher's position may have shaped what participants said and what was done to test it; where an alternative reading is available and how it was examined; where the account depends on a small number of participants. A threat named without an instantiation is a textbook list, and a threat named with no action is an admission. Each entry gets one of three outcomes: addressed by design, addressed by analysis, or unaddressed and carried into the limitations with its effect on the claims stated. Unaddressed is acceptable and often honest; silently listing it is not. *Correct result: a threat register with each threat instantiated, its treatment, and where an unaddressed threat bites on which conclusion.*

**Step 5. Make qualitative rigour a set of practices, not a vocabulary.**
The criteria commonly invoked (credibility, transferability, dependability, confirmability, or the equivalent set in the tradition used) mean nothing unless the practices behind them are evidenced. The audit trail is a dated record of analytic decisions, code frame versions and why categories changed, sufficient for another researcher to follow the reasoning from data to conclusion; it is built during analysis and cannot be reconstructed afterwards. Negative case analysis means actively searching for data that contradicts the developing account and reporting what it did to that account, the strongest single rigour practice available and the least often done. Thick description means enough contextual detail for a reader to judge transferability to their setting, which is not the same as long quotations. Member checking must be used appropriately: participants can validate that they were understood, and cannot validate an analytic interpretation operating above their own account, a distinction that is contested and must be handled explicitly rather than claimed as blanket validation. Peer debriefing and independent coding are evidence when disagreements and their resolutions are reported, and decoration when only an agreement figure appears. State which practices were used, where the evidence sits, and which were not used and why. *Correct result: a rigour table mapping each claimed criterion to the specific practice, the artefact that evidences it, and its location.*

**Step 6. Write reflexivity as analysis, not confession.**
A reflexivity statement listing the researcher's demographic characteristics and expressing awareness of bias does nothing, and examiners increasingly say so. Analytic reflexivity does three things. It identifies the specific ways the researcher's position plausibly shaped this study: what access it opened and closed, what participants were likely to say or not say to this person, which interpretations came easily. It reports what was done about each: a probe added, an alternative reading tested, a co-coder brought in where the researcher's prior was strongest, a participant group deliberately sought whose perspective the researcher was least likely to anticipate. And it reports what changed as a result, because reflexivity with no consequence is a performance. Statements about the researcher's relationship to the field belong here too, particularly insider research, where the analytic question is what the insider knows that an outsider would miss and what an insider cannot see. *Correct result: a reflexivity section in which every stated position is followed by a specific consequence for the study and a specific action taken.*

**Step 7. Defend the sampling logic and state the stopping rule.**
Sampling is justified against purpose, not size. For probability designs, defend the frame's coverage, the selection mechanism, the achieved versus target sample, the response rate and non-response risk, and the basis of any power calculation including the effect size assumed and where it came from, since a power analysis built on an effect size chosen to produce a feasible sample is circular. For purposive designs, name the strategy, the criteria, and why this composition can address the question. For theoretical sampling the stopping rule is the hard part, since saturation is frequently claimed and rarely evidenced. The defensible version specifies in advance what saturation means here (no new codes, no new properties of existing categories, or no change to the developing theory, which are three different thresholds), records where new data stopped changing the account, and reports the evidence: the interview after which no new categories appeared, and how many further interviews confirmed it. *Correct result: a sampling justification tied to purpose, and a stopping rule stated in advance with the evidence that it was met.*

**Step 8. Establish instrument validity with evidence, not construction.**
Where a validated instrument is used, report the validation evidence and, critically, the population it was validated in: validation does not transfer automatically, and adapting an instrument by translating, shortening, changing the response scale or changing the referent breaks it and requires re-evidencing. Where an instrument was built for this study, report the construction process against the construct definition, the content validation, the pilot evidence, the factor structure or equivalent, and reliability estimates with their bases. Where a construct is measured with a single item, say so and state the consequence (K4 §7). For qualitative instruments the equivalent argument concerns the interview guide or observation protocol: how it was derived from the research questions and the framework, how it was piloted, and how it changed. *Correct result: a validity argument per instrument, naming the evidence and the population it comes from, with every adaptation flagged.*

**Step 9. Report what the pilot changed.**
A pilot reported as having been conducted, with no consequences, is evidence that it was not used. Report what was tested, what problems it revealed, what was changed as a result, and what was deliberately not changed and why. Typical yields worth reporting: questions that were understood differently from intended, a routing error, an interview guide whose opening question closed participants down, a recruitment approach with a lower yield than assumed, timing that exceeded what participants would tolerate, and a construct that turned out to be locally meaningless. Also state whether pilot data was included in the main analysis, and if so on what basis, since including data collected with a different instrument is a defect that has to be argued for rather than glossed. *Correct result: a before-and-after account of the instrument and procedures, with the pilot's yield stated and the inclusion decision explicit.*

**Step 10. Build the transparency and reproducibility position.**
Decide, and state, what is available and what is not: the instrument, the coding frame, the analysis code, the data, the protocol. For quantitative work the strong position is that another researcher with the data and the code can reproduce every number. For qualitative work full data sharing is often impossible for confidentiality reasons, and the equivalent commitment is procedural transparency: the guide, the coding structure, worked examples of the analytic steps, and the audit trail available to an examiner. Where data cannot be shared, say why and say what can be. Handle ethics and consent constraints explicitly rather than using them as a general excuse, since consent that did not anticipate sharing is a real constraint and an unexamined assumption that sharing is impossible is not. *Correct result: an availability statement covering data, code, instruments and protocol, with reasons for each restriction.*

**Step 11. Decide on pre-registration and, where used, report deviations.**
Pre-registration is expected in some fields, growing in others and unusual in several, and the decision should rest on what it buys: it credibly separates confirmatory from exploratory analysis, which is otherwise almost impossible to demonstrate after the fact. Where used, the registration must be specific enough to constrain: hypotheses, primary outcome, analysis, exclusion rules and sample size rule. A registration that leaves the analysis open constrains nothing. Then report every deviation with its reason and likely effect, because an unreported deviation discovered by a reviewer or examiner damages credibility more than the deviation ever would. Where pre-registration is not used, the honest alternative is to label analyses confirmatory or exploratory in the write-up and hold to it. *Correct result: a stated position on registration with reasons, and a complete deviation log where it applies.*

**Step 12. Convert every unaddressed weakness into a limitation with a consequence.**
Write the limitations section from the threat register, the rejected alternatives, the sampling constraints and the instrument compromises, not from a generic template. Each limitation names what it affects, how, and in which direction where that is knowable. "The sample was drawn from one region" is a fact; "the sample was drawn from one region, which limits the claim to regional rather than national and may overstate the effect because this region has unusually high service provision" is a limitation. Then check upward: does every conclusion respect these limitations, or does the discussion quietly reclaim the territory the methodology conceded? That inconsistency is one of the most common findings in an examination. *Correct result: a limitations section derived from the study's own decisions, and a check that no conclusion exceeds it.*

## 8. Analytical framework

Every methodological choice is defended on the same structure:

    Research question requirement → Options available →
    Comparative evaluation against the requirement →
    Choice with the rejection reason for each alternative →
    Cost of the choice → Effect of that cost on the claims → Mitigation or limitation

The link that distinguishes a doctoral chapter is *Comparative evaluation*: without it there is description, with it there is argument. The link most often broken is *Effect of that cost on the claims*: candidates concede a limitation in chapter three and write conclusions in chapter seven that ignore it.

A second structure governs coherence, and it must hold at every level:

    Ontology → Epistemology → Methodology → Methods → Analysis → Claim type

A break anywhere invalidates everything below it. The commonest breaks are between epistemology and analysis (a constructivist position with a frequency-counting analysis), and between analysis and claim type (an interpretive analysis producing a generalisation about a population). Test the chain by asking at each link whether the level below follows from the level above, and whether it would follow equally from a different level above, which detects a position that is doing no work.

## 9. Output format

**1. Design requirements table.**

| Research question | Claim type | Unit and level | Temporal requirement | Population to reach | Precision needed |
|---|---|---|---|---|---|

**2. Comparative justification, per major choice.**

| Choice point | Option | What it delivers | What it costs | Rejection reason and its category | Chosen? |
|---|---|---|---|---|---|

**3. Philosophical position and coherence check.** The position in the candidate's words, the forward and backward checks, and any acknowledged tension with its argument.

**4. Validity threat register.**

| Threat | Instantiated in this study as | Addressed by design / analysis / unaddressed | Evidence | Which conclusion it bites on |
|---|---|---|---|---|

**5. Rigour practices table** (qualitative or mixed).

| Criterion claimed | Practice used | Artefact evidencing it | Location in thesis | Not used, and why |
|---|---|---|---|---|

**6. Reflexivity statement.** Position, consequence, action, and what changed.

**7. Sampling justification** with the stopping rule stated in advance and the evidence it was met.

**8. Instrument validity argument**, including the validation population and every adaptation.

**9. Pilot report.** Tested, found, changed, not changed, and the inclusion decision.

**10. Transparency statement.** Data, code, instruments, protocol: what is available, what is not, and why.

**11. Deviations from protocol or plan**, complete, with reasons and effects.

**12. Limitations**, derived from the above, each with its consequence for a named claim.

**When the evidence is thin, the format must not force fabrication (K4 §1).** Where an alternative was never actually considered, the honest entry says the choice was made by default or by constraint, which is a defensible answer and a far better one than a retrospective rationalisation an examiner can smell. Where a rigour practice was not used, the row says not used with the reason, and the criterion is not claimed. Where saturation was not systematically assessed, it is not claimed; the sample is defended on other grounds.

## 10. Quality checks

Run before the chapter is submitted or defended. These sit on top of K4 §8.

1. Is every major choice defended against named alternatives rather than described?
2. Does the set of alternatives include the one an expert examiner would most plausibly have chosen?
3. Is each rejection reason categorised as validity, feasibility, ethics, access or paradigm fit, and is any feasibility rejection dressed as a principled one?
4. Does the philosophical position imply these methods, and would the methods imply this position if read backwards?
5. Is every validity threat instantiated in this specific study, with a named treatment or an explicit admission?
6. Does every rigour criterion claimed have a practice and an artefact behind it, at a stated location?
7. Is the reflexivity statement analytic, with a consequence and an action for every position stated?
8. Is the sampling stopping rule stated in advance, and is the evidence that it was met reported?
9. Is any saturation claim evidenced rather than asserted?
10. Is the validation population of every instrument reported, and is every adaptation flagged as breaking the original validation?
11. Does the pilot report say what changed, and is the pilot-data inclusion decision explicit?
12. Are all deviations from the plan or registration reported, including the inconvenient ones?
13. Does every limitation state a consequence for a named claim rather than a fact about the study?
14. Does any conclusion elsewhere in the thesis exceed what this chapter concedes?
15. Could a competent researcher in the field reproduce the study from this chapter alone?

## 11. Common failure modes

| Failure | How to recognise it | How to prevent it |
|---|---|---|
| **Descriptive methodology** | The chapter says what was done, never why not otherwise | Build the comparative table (Step 2) |
| **Straw alternatives** | Rejected options are obviously unsuitable | Include the examiner's likely choice (Step 2) |
| **Decorative paradigm** | The philosophical section could be deleted without affecting the methods | Run the backward coherence check (Step 3) |
| **Textbook threat list** | Threats named in general terms with no instantiation | Instantiate or delete each (Step 4) |
| **Rigour by vocabulary** | Trustworthiness criteria claimed, no practices visible | Map each criterion to an artefact (Step 5) |
| **Confessional reflexivity** | Demographics plus an expression of awareness | Require consequence and action (Step 6) |
| **Asserted saturation** | "Saturation was reached" with no rule and no evidence | State the rule in advance and evidence it (Step 7) |
| **Borrowed validation** | An instrument validated elsewhere used unchanged in a different population | Report the validation population and flag adaptations (Step 8) |
| **Circular power analysis** | The effect size assumed is the one that makes the sample feasible | Source the effect size independently (Step 7) |
| **The consequence-free pilot** | A pilot is reported and nothing changed | Report what changed and what did not (Step 9) |
| **Silent deviation** | The analysis differs from the plan and the difference is unmentioned | Log every deviation with its effect (Step 11) |
| **Reclaimed territory** | Chapter three concedes what chapter seven ignores | Check conclusions against limitations (Step 12) |
| **Retrospective rationalisation** | Every choice was apparently the ideal one for principled reasons | Record decisions with dates during design (Step 2) |
| **Generic limitations** | Limitations that would fit any study | Derive them from this study's own decisions (Step 12) |

## 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 methodology chapter is the chapter a viva examines line by line, and a candidate who cannot explain why an alternative was rejected will be found out within minutes. The skill's job is to produce the questions and the gaps, not the prose.

2. **Never invent a justification for a choice that was made by constraint.** If the design was chosen because it was the only one the access permitted, that is the reason, and it is an acceptable one. Manufacturing a principled rationale after the fact is a fabrication about the research process (K4 §2.5) and it is detectable, because a constructed rationale is always slightly too tidy.

3. **Never describe a procedure that was not carried out.** Piloting, member checking, independent coding, a reflexive journal and an audit trail are either done or not. Writing them into a chapter because the format expects them is fabrication of methodology, and it is the specific form most likely to be exposed when an examiner asks to see the artefact.

4. **Never report a validation, a reliability figure, a response rate or a saturation point that was not obtained.** Where a figure is not available, the chapter says it is not available.

5. **Never assign a philosophical position to the candidate.** The position must be one they hold and can defend under questioning. The skill may explain what positions commit their holders to, and may identify incoherence between a declared position and the methods used, and must not choose.

6. **Never present one paradigm as more rigorous than another.** Quantitative and qualitative traditions have different criteria, different threats and different warrants, and treating one set as the standard against which the other is measured is a category error that damages both. Where criteria genuinely conflict across traditions, present the disagreement.

7. **Never soften an unaddressed threat into a mitigated one.** "This was partially addressed by" is the phrasing that converts an honest admission into an overclaim. If nothing was done, the register says nothing was done, and the limitation carries the consequence.

8. **Never generate methodological literature citations from model knowledge.** Methods literatures are heavily cited and heavily misattributed, and plausible attributions of a criterion, a threat taxonomy or a design label to the wrong originator are a standard AI failure with real consequences in an examination (K4 §2.4).

9. **Where a design cannot answer a question, say so rather than producing a better defence of it** (K4 §1). A persuasive justification for an inadequate design increases the harm, because it delays the discovery until examination.

## 13. Best-practice principles

- **Justification is comparative or it is decoration.** The sentence that makes a methodology chapter doctoral is "X was rejected because", repeated for every real alternative.
- **Record decisions when you make them.** A dated design log converts the whole chapter from reconstruction into report, and it is the difference between a defence that sounds honest and one that sounds assembled.
- **Answer the examiner's alternative directly.** In every field there is an obvious other way to have done the study. If your chapter does not name it, the examiner will, and you will be answering it for the first time in the room.
- **A position that does no work is worse than no position.** An examiner who finds the philosophical section detachable will conclude the candidate does not understand it, which is usually correct.
- **Threats are specific or they are furniture.** "Social desirability bias" is a category. "Participants were asked about adherence by the clinician responsible for their care" is a threat.
- **Rigour is a set of things you did, and each leaves an artefact.** If there is no artefact, the claim cannot be evidenced, and an examiner asking to see it is entitled to.
- **Negative case analysis is the highest-yield qualitative rigour practice and the least often done.** Actively hunting for the data that breaks your account improves the account every time and defends it permanently.
- **Reflexivity without consequence is performance.** The question is not who you are; it is what difference it made and what you did about it.
- **Saturation is a claim about your data, not a ritual.** State the threshold, record where it was reached, and report the confirming interviews.
- **Every limitation should be traceable to a decision you made.** A limitations section that could belong to any study signals that the specific ones have not been identified.
- **Concede in the methodology and hold the concession in the conclusion.** Reclaiming ground later is the inconsistency examiners find most easily, because they read the two chapters against each other deliberately.
- **Transparency is a design decision, not a formality.** Deciding at the start what will be shareable shapes the consent, the anonymisation and the file structure, and it is nearly impossible to retrofit.

## 14. Worked example

Generic fictional scenario, academic, health services research.

**INPUT**

A doctoral candidate has completed 28 semi-structured interviews with community pharmacists about how they handle prescribing queries, analysed thematically. Her methodology chapter declares a critical realist position, states that trustworthiness was ensured through credibility, transferability, dependability and confirmability, reports that saturation was reached at interview 24, and describes the interview procedure in detail. Her supervisor says the chapter is descriptive and will not survive.

**PROCESS**

*Step 1.* Requirements are derived per question. Two of her three questions require interpretive access to reasoning, which interviews suit. The third asks how frequently a particular type of query is escalated, which interviews cannot answer, and which she had been answering from interview recall. That mismatch is caught here rather than at the viva.

*Step 2.* The comparative table is built. For the design, the alternatives were observation of practice, a document study of escalation records, a survey, and the chosen interviews. Observation would have given behaviour rather than accounts, and was rejected on access and on patient confidentiality, which is a legitimate reason and is named as an access-and-ethics rejection rather than a principled one. The document study would have answered question three properly, and was rejected because records were not available across sites, which is now stated openly and carried into the limitations for question three.

*Step 3.* Coherence. The declared critical realist position is checked backwards. Her analysis stops at describing patterns in what pharmacists said, never proposing generative mechanisms operating beneath them, which is precisely what critical realism commits her to. Two options: change the position, or do the analytical work the position requires. She chooses the second, and a fourth analytic pass looks for mechanisms, which materially improves the findings chapter.

*The judgement call.* The saturation claim. Her records show she stopped at 28 because her funding period ended. She had written "saturation was reached at interview 24" because the literature said that is what one writes. The temptation is to keep the sentence, since nobody can disprove it. The decision taken is to remove it and replace it with what actually happened: the coding record shows the last new category appearing at interview 21, with seven subsequent interviews adding properties to existing categories but no new categories, and the sample was in any case bounded by the funding period. That is a weaker-sounding and far stronger claim, because it is evidenced and because it concedes the real constraint. An examiner who asks how saturation was assessed now gets an answer with a document behind it.

*Steps 5 and 6.* The four trustworthiness criteria are mapped to practices. Credibility has independent coding of six transcripts, and the disagreements and their resolution are now reported rather than only the agreement level. Dependability has an audit trail: her memos exist, and they are organised into an appendix. Confirmability is weak, so a negative case search is run retrospectively across the transcripts, which finds three pharmacists whose accounts contradict the main theme and who had been under-represented, and the theme is revised. Transferability is addressed by adding site-level contextual description. Reflexivity is rewritten: she is a former community pharmacist, and the analytic consequence is that participants used professional shorthand freely with her, which gave access an outsider would not have had, and that she initially coded escalation as a failure of confidence because that was her own framing, which the negative cases corrected.

**OUTPUT**

A chapter that argues rather than describes, concedes that question three cannot be answered by this design and says so, evidences four rigour criteria against artefacts, replaces an unevidenced saturation claim with a documented sampling account, and carries three specific limitations into the discussion with their consequences named.

`RESEARCHER SIGN-OFF REQUIRED` on the decision to retain question three with an explicit statement that the design cannot answer it, rather than dropping it (K5 §2.7).

## 15. Advanced usage

**Defending a mixed-methods design.** The justification burden is higher because the paradigm question is sharper: the chapter must state the position on whether the traditions are commensurable, name the design type and the sequencing logic, and specify the integration point where the strands actually meet. Two strands reported side by side with no integration is a common and easily attacked weakness. See **15.17 Longitudinal and Multi-Study Design**.

**Defending a design that changed mid-study.** Access failures, a pandemic, a refused site, a withdrawn dataset. The strong version reports the original design, the event, the options at that point, the decision and the reasoning, and the effect on the claims. This reads as competent research management. The weak version presents the revised design as though it were always the plan, and it is usually detectable from the research questions, which will still be shaped by the original.

**Insider and practitioner research.** Where the researcher works in the setting studied, the methodology chapter carries additional burdens: the effect of existing relationships on what participants say, the power asymmetry where the researcher has organisational authority, the ethical handling of information obtained outside the study, and the analytic question of what an insider cannot see. Handled explicitly, insider position is a strength; handled as a disclosure, it is an attack surface.

**Pre-registration for qualitative and exploratory work.** The confirmatory-exploratory distinction transfers imperfectly. What can usefully be registered in advance is the question, the sampling strategy and stopping rule, the analytic approach, and the reflexive position, leaving the analysis itself open. Registering something is generally better than registering nothing, provided what is registered is not so vague that it constrains nothing.

**Preparing the chapter for examination.** Run the whole chapter as an adversarial read: for each paragraph, write the question an examiner would ask, and answer it. Choices with no answer are the ones to fix now. This hands directly to **15.19 Examiner-Proofing and Defence Strategy**.

## 16. Skill chain

**Recommended previous skills:**
- **01.04 Research Method Selection.** Hands over the chosen design and the options considered, which this skill converts into a defence.
- **15.13 Doctoral Research Positioning and Originality.** Hands over the sized contribution the design must be capable of supporting.
- **15.15 Theory Building and Conceptual Contribution.** Hands over propositions that specify what the design must test.
- **15.09 Ethics Clearance Application.** Hands over the ethical constraints that legitimately shaped the design.

**Recommended next skills:**
- **15.17 Longitudinal and Multi-Study Design.** Where the requirements table shows one design cannot serve all the questions.
- **15.10 Data Analysis Chapter Development.** Takes the justified design into the reporting of results.
- **15.18 Thesis Architecture and Chapter Coherence.** Checks that the concessions made here hold across the whole thesis.
- **15.19 Examiner-Proofing and Defence Strategy.** Turns the threat register and rejected alternatives into examination preparation.

**Runs well alongside:**
- **05.01 Statistical Analysis Planning**, for the analytical decisions this chapter must support.
- **02.07 Scale and Measurement Selection**, for the instrument validity argument.
- **13.01 Research Quality Review**, which audits the design independently of its findings.

---
A Yazi Supplied Skill and resource.