---
name: discussion-and-contribution-development
description: >
  Builds the discussion chapter: answering each research question explicitly,
  interpreting findings against the literature including where they disagree,
  stating a contribution calibrated to masters level rather than overclaimed,
  distinguishing theoretical from practical implications, interpreting
  unexpected and null findings, writing specific limitations, and a conclusion
  that answers the question the introduction asked. Use for "discussion
  chapter", "chapter 5", "what is my contribution", "my findings contradict the
  literature", "how do I write limitations", "future research section",
  "conclusion chapter", "am I overclaiming".
category: 15 Academic University Research
ref: "15.11"
tier: 3
inherits: [K2, K3, K4, K5]
---

# Discussion and Contribution Development

## 1. One-line description

A method for turning results into an argument: answering each research question explicitly, situating every finding against the literature including where it disagrees, stating a contribution at the level a masters degree actually supports, and closing with a conclusion that answers the question the introduction asked.

## 2. What this skill is used for

**The research problem it solves.** The discussion is where the most marks are available and where the most are lost, because it is the only chapter that requires argument rather than execution. Five failures recur. **Restatement**: the chapter walks through the results again in prose, adding nothing, which markers describe as a second results chapter. **Unanswered questions**: the research questions posed in the introduction are never explicitly answered, so the reader has to assemble the answer themselves. **Disagreement mishandled**: a finding contradicts prior work and the student either buries it, apologises for it, or attributes it to their own error, when contradiction is a finding and often the most interesting one available. **Overclaiming**: a study of 40 people in one organisation is described as demonstrating something about a sector, or as filling a gap in the literature, and overclaiming is penalised more heavily than modest claiming because it shows a failure of judgement rather than a limitation of scope. **Generic limitations**: "small sample size, further research is needed" appears in tens of thousands of dissertations and earns nothing in any of them, because it names no specific consequence for any specific finding.

**Where it sits in the research lifecycle.** After the results chapter has established what was found, and before or alongside the conclusion. It is the chapter that connects back to the literature review and the framework, closing the loop the introduction opened.

**Typical use cases.**
- Writing the discussion from a completed results chapter.
- Working out what the study's contribution actually is, and at what level to claim it.
- Handling a finding that contradicts established work.
- Interpreting an unexpected or null result rather than apologising for it.
- Replacing a generic limitations list with specific, consequential ones.
- Writing future research that follows from this study rather than from the topic.
- Writing a conclusion that answers the introduction's question.
- Repairing a discussion that restates results.

**Who uses it.** Masters candidates writing chapter five; supervisors reviewing a discussion draft; students in every tradition, since the requirement to interpret against the literature and state a contribution is common to all of them.

## 3. When to use it

- The results chapter is complete and stable, and the findings will not change.
- A supervisor has said the discussion restates the results.
- You cannot state your contribution in one sentence.
- A finding disagrees with prior work and you do not know how to handle it.
- An analysis produced nothing and you do not know what to say about it.
- Your limitations section is generic and you know it.
- The future research section reads like a list of everything adjacent to your topic.
- You need to check whether the conclusion answers the question the introduction asked.

## 4. When NOT to use it

- **The results are not final.** A discussion built on results that then change has to be rewritten entirely, because every interpretation is anchored to a specific number or theme. Finish **15.10 Data Analysis Chapter Development** first, including the assumption checks, since a test that has to be replaced changes what there is to discuss.
- **What is needed is the presentation of results.** Reporting findings, statistics, themes and tables is **15.10**, and this skill assumes that work is done. In the other direction: 15.10 deliberately excludes all interpretation, literature comparison, implication and limitation-by-finding, and hands every one of them here.
- **The literature review is thin or was never properly done.** You cannot situate findings against a literature you have not read, and a discussion that compares results to three papers found afterwards is visibly retrofitted. Go back to **15.06 Systematic Literature Review** or **15.02 Academic Literature Search Strategy**. The discussion's quality is capped by the literature review's coverage.
- **The claim being developed is a genuinely original theoretical contribution.** Constructing new theory with defined constructs, propositions, mechanisms and boundary conditions is doctoral work and is **15.15 Theory Building and Conceptual Contribution**. Establishing and defending originality at doctoral level is **15.13 Doctoral Research Positioning and Originality**. This skill calibrates to a masters contribution, which is real and modest, and claiming beyond it is the failure this skill exists to prevent.
- **The output is a practitioner or client report.** Turning findings into organisational recommendations, with feasibility, ownership and sequencing, is **08.04 Recommendation Development** and **12.05 Executive Research Reporting**. An academic discussion argues about knowledge; a practitioner report argues about action, and the two are marked by opposite criteria.
- **A finding cannot be interpreted because the design cannot support the interpretation.** Where the data are cross-sectional and the interesting reading is causal, or where the sample cannot represent the population the interpretation would concern, the correct output is the limited interpretation plus a statement of what design would license the fuller one. Writing the fuller interpretation with a hedge attached is overclaiming with a disclaimer (K4 §3.2, §3.3).
- **Your institution prohibits AI assistance for this task.** See §12 item 1. This chapter is where a candidate's own analytical judgement is most directly assessed, and where borrowed argument is most visible in a viva.

## 5. Required inputs

**Required.**
- **The final results, as reported.** Every interpretation must be anchored to a specific finding with its base and its strength, and interpretations drafted against remembered results drift from what the chapter actually says.
- **The research questions, in their introduction wording.** Each one must be answered explicitly, in the words it was asked in, or the reader cannot see that it was answered.
- **The literature review, in detail.** Specifically: which studies found what, in which populations, with which methods, so that agreement and disagreement can be located precisely rather than in general terms.
- **The conceptual or theoretical framework.** Findings are read back against it, including the parts it failed to account for.

**Optional, and what each one adds.**
- **The gap statement from the literature review or the proposal.** The contribution is best stated as what the gap now contains, and having the original wording prevents the contribution drifting into a claim about a different gap.
- **The marking rubric's contribution and discussion criteria.** Tells you what the department is actually looking for, and how many marks depend on the contribution being explicit.
- **Studies published since the literature review was written.** A discussion that ignores work published during the study reads as dated, and a single recent paper agreeing or disagreeing materially strengthens the chapter.
- **The methodology chapter's limitations section.** The discussion's limitations build on those rather than repeating them, and having both prevents duplication.
- **Notes on what surprised you during fieldwork and analysis.** These are usually the discussion's most valuable material and they are almost always lost if not written down at the time.

## 6. Questions to ask before starting

1. **What did the introduction promise the study would do?** The conclusion must deliver against that promise in its own terms. Default if unanswered: extract the promise from the introduction's aim and objectives, and if the study did something different, say so explicitly rather than quietly answering a different question.
2. **What does this department expect a masters contribution to be?** Expectations vary from "demonstrates competence in applying an established method" to "produces a defensible novel finding". Default: claim an empirical or contextual contribution, which almost every competent masters study genuinely has, rather than a theoretical one, which few do.
3. **Which findings agree with prior work, which disagree, and which have no prior work to compare against?** Determines the shape of the chapter. Default: build the convergence and divergence map before writing, since it is the chapter's structure.
4. **Are the discussion and conclusion separate chapters or one?** Determines where the answers to research questions land and where limitations and future research go. Default: check the departmental structure, and if separate, put the interpretation in the discussion and the summative answer, contribution, limitations and future research in the conclusion.
5. **What did the framework fail to account for?** This is where a masters contribution most often lives and it is the question students least often ask. Default: identify the residual explicitly, because a study whose framework explained everything has usually not looked hard enough.
6. **What is the single most defensible thing this study established?** If you cannot answer, the contribution statement will be vague. Default: draft it as one sentence with its population and its bounds attached, then test it against the results.
7. **Where would a hostile reader attack the interpretation?** Determines what must be pre-empted. Default: assume the attack is on generalisation, on the causal reading, or on an alternative explanation you have not addressed, since those are the three most common.

## 7. Step-by-step methodology

**Step 1. Answer each research question explicitly, in one sentence, before writing anything else.**
Take the research questions in their introduction wording and write a direct answer to each, in a single sentence, with its evidence reference and its bounds. "RQ2 asked whether the association differed by attainment band. It did not: the difference between bands was not statistically detectable in this sample, and the interval leaves open effects of a size that would matter educationally." These sentences are the chapter's skeleton and they usually become its section openers. Two failure states show up immediately at this step: a question you cannot answer from the results, which means either the results are incomplete or the question was never answerable and that must be said; and a question whose answer is "partly", which is legitimate and must be written as a bounded answer rather than allowed to blur. Write the answer even when it is unwelcome. *Correct result: one direct sentence per research question, each anchored to a specific result.*

**Step 2. Build the convergence and divergence map against the literature.**
For each substantive finding, tabulate: the finding, the studies that agree, the studies that disagree, the studies that are silent, and the closest comparable population and method. This table is the chapter's analytical core and it prevents the two commonest structural failures, which are discussing findings in isolation from the literature and discussing the literature in isolation from the findings. Do it for every finding, including the null ones, and specifically look for the third category, where nothing in the literature speaks to your finding at all, since that is where a contextual contribution usually sits. Be precise about what agreement means: a study finding the same direction in a different population with a different measure is partial agreement, and calling it confirmation overstates it. *Correct result: a finding-by-finding map naming specific studies in each column, with the comparability of each stated.*

**Step 3. Diagnose disagreement before explaining it, and never apologise for it.**
Where a finding contradicts prior work, work through the possible causes in order before concluding: **population** (different people, different context, different sector or country); **measurement** (different instrument, different construct definition, different question wording, stated versus observed behaviour); **method** (different design, different analysis, different time horizon); **temporal** (the phenomenon genuinely changed, or the earlier study preceded a structural shift); **analytical** (different covariates, different handling of a confounder, different cut-points); and only after all of these, **genuine disagreement** (comparable studies, incompatible results). The first five are more common than the sixth. Then write the disagreement as a finding: state both positions, state the diagnosed cause, state which is better evidenced for which population, and where the disagreement is genuine and unresolved, say so. What must not happen: attributing the disagreement to your own study's inadequacy by default, which is the reflex of an anxious writer and is usually unjustified, since your study may be the better-designed one for this population; quietly omitting the contradicting literature, which is a completeness failure (K4 §4.1); or splitting the difference. A well-handled disagreement is frequently the strongest passage in a masters dissertation, because it demonstrates exactly the judgement being assessed. *Correct result: every disagreement named, diagnosed against the six causes, and resolved with reasons or explicitly left open.*

**Step 4. Read the findings back against the framework, including what it failed to explain.**
For each element of the conceptual framework: was it supported, partly supported, unsupported, or not testable with the data obtained? Then, more importantly, what appeared in the data that the framework had no place for? In quantitative work this is the variance the model does not account for and the pattern that sits outside the specified relationships; in qualitative work it is the residual material that did not map to the sensitising concepts. Report it, because the residual is where a masters-level contribution most often lives: it is the point at which your specific context exceeded the borrowed theory. Where the framework performed poorly overall, say so plainly and consider what that implies about the theory's boundary conditions in this context, which is a legitimate and modest theoretical observation. *Correct result: a per-element verdict on the framework, plus an explicit account of what it did not accommodate.*

**Step 5. State the contribution explicitly, and calibrate it to what a masters study supports.**
Contribution must be stated in its own passage, under its own heading, in plain declarative sentences. Do not leave the reader to infer it. Use the contribution types and pick the ones that genuinely apply: **empirical** (new evidence about a phenomenon, the most common and most defensible masters contribution); **contextual** (established relationships examined in a population, sector, country or setting where they have not been examined, which is a real contribution and is frequently undersold by students who think it does not count); **methodological** (an instrument adapted and its performance reported, a procedure applied in a new setting, a measurement problem exposed); **theoretical** (a boundary condition identified, a construct refined, a framework found inadequate in a specified way, which at masters level is modest and specific rather than the construction of new theory); and **practical** (implications for a defined group of practitioners, which counts in applied disciplines and must be evidenced rather than asserted). Then calibrate, with three tests. The **scope test**: does the claim extend beyond the population, period and setting the sample supports? The **novelty test**: could you name the specific prior work this adds to, and say what it adds? If not, the claim is too vague to be a contribution. The **subtraction test**: if this study had not been done, what specifically would not be known? A contribution that survives subtraction is real. Write the contribution with its bounds attached in the same sentence, since an unbounded contribution claim is the single most common overclaim: "this study provides the first evidence, in this sector and country, that..." is defensible; "this study demonstrates that..." usually is not. **Overclaiming is penalised more heavily than modest claiming**, because a modest claim shows judgement and an inflated one shows its absence. `RESEARCHER REVIEW RECOMMENDED` on the contribution statement, since judging what is genuinely novel in a field requires knowledge of that field's current state (K5 §2.1). *Correct result: a contribution section naming the type or types, stating what specifically is now known that was not, with the bounds in the sentence.*

**Step 6. Separate theoretical from practical implications, and evidence both.**
These are different claims for different audiences and merging them weakens both. **Theoretical implications** concern what the findings mean for the concepts, relationships and theories in the literature: a boundary condition on an existing theory, a construct that behaves differently in this context, a mechanism that the data suggest operates differently than assumed, a relationship whose sign or strength differs. Each must be tied to a specific finding and stated at the strength that finding supports. **Practical implications** concern what the findings mean for people who act: who could do what differently, on the basis of what finding, and under what conditions. The discipline here is to derive them rather than assert them: an implication that does not follow from a specific finding is the author's opinion, and markers identify it easily. Constrain them by the evidence: a study of one organisation supports implications for that organisation and a stated argument about similar ones, not a set of sector recommendations. And state the confidence of each implication, capped by the finding beneath it (K3 §3.6). *Correct result: two distinct subsections, each implication traceable to a named finding, each bounded and confidence-marked.*

**Step 7. Interpret the unexpected and the null rather than apologising for them.**
An unexpected finding is an opportunity and students routinely treat it as an embarrassment. Interpret it in three moves: state what was expected and why (from the framework or the literature), state what was found, and then work through the candidate explanations, which are the same six as in Step 3 plus the possibility that the expectation was wrong. Say which explanation you find most plausible and why, at the confidence the evidence supports, and say what would distinguish between the remaining candidates. Null results need the same treatment plus one more distinction that students routinely miss: **failing to detect an effect is not the same as establishing there is none**, and which of these you can claim depends on the precision of the estimate. Where the confidence interval is wide and includes effects that would matter, the honest statement is that the study was not powerful enough to settle the question, which is a statement about the study; where the interval is narrow and close to zero, you have evidence of a small or absent effect, which is a statement about the world and is a genuine contribution. Never write a null result as a failure, never attribute it to the sample size without checking whether the interval supports that reading, and never quietly convert it into a marginal positive with softening language. *Correct result: each unexpected and null finding interpreted with candidate explanations, a stated most-plausible reading at an appropriate confidence, and for nulls an explicit statement of which claim the precision supports.*

**Step 8. Write limitations that are specific, consequential and directional.**
A limitation has three parts, and a generic list has none of them: what the limitation is, which specific findings it affects, and in which direction it would bias them. "The sample was small" is a fact. "The sample of 40 gives limited precision on the subgroup comparison in RQ2, where the interval spans effects from negligible to substantial, so that analysis should be read as inconclusive rather than negative" is a limitation. Work through the categories systematically and keep only those that bite: sampling and who is missing from the sample, with the direction the absence would push results; measurement and what the instrument does not capture of the construct; design and what it cannot establish, particularly temporal order and causal direction for cross-sectional work; analysis and the judgement calls made; context and period, and what would make the findings expire; researcher position and its effect on what participants would say; and response or attrition and who is likely to have dropped out. Distinguish these from the methodology chapter's design limitations, which were known in advance: this chapter's limitations are those visible in the light of what was actually found. State mitigations where they exist and do not over-mitigate, since a limitation talked away is a limitation the examiner will raise. And do not list limitations that are not limitations of this study, which is a form of padding examiners notice. *Correct result: four to eight limitations, each naming the affected finding and the direction of the effect, with none generic.*

**Step 9. Write future research that follows from this study specifically.**
The test is whether each suggestion could have been written without doing your study. If it could, delete it. "Further research with a larger sample" fails the test; "a longitudinal design measuring the same cohort at intake and at twelve months would establish whether the association observed here reflects selection into the programme or change produced by it, which this cross-sectional design cannot separate" passes it, because it follows from a specific limitation of a specific finding. Derive suggestions from three sources: the questions your findings opened, the alternative explanations you could not rule out, and the residual your framework could not accommodate. For each, state the question, the design that would answer it, and what it would settle. Three or four well-specified suggestions are worth more than a list of twelve. *Correct result: three to five suggestions, each traceable to a specific finding, limitation or residual in this study, each with a design attached.*

**Step 10. Write the conclusion to answer the introduction's question.**
The conclusion is not a summary of the dissertation and it is not a second discussion. It does four things: it answers the question the introduction asked, in the introduction's own terms, in a short and direct passage; it states the contribution; it states the principal limitations in one compressed paragraph; and it closes with what follows. Read the introduction immediately before writing it, because the commonest defect is a conclusion that answers a question slightly different from the one posed, which the examiner notices because they read both. No new evidence, no new citations, no new claims appear here. If the study did not answer its question, the conclusion says what it did establish and why the original question remains open, which is a legitimate and respectable outcome and considerably better than a conclusion that pretends. *Correct result: a conclusion that a reader could compare line by line with the introduction and find answered.*

**Step 11. Run the overclaiming audit.**
Read the chapter with one purpose: find every claim that exceeds its evidence. Check specifically for **population drift** (findings about your sample described as findings about a sector, a profession or a country); **causal drift** (associations described with the vocabulary of cause: drives, leads to, results in, impacts, per K4 §3.2); **significance without testing** (differences described as notable or clear where no test was run, per K4 §3.1); **confidence inflation** (a moderate finding producing a confident implication, per K3 §3.6); **contribution inflation** (a competent study described as filling a gap, transforming understanding or challenging established theory); and **temporal drift** (a finding from one period stated as a permanent characteristic). Then run the reverse check, which is equally important and less often done: is any strong finding hedged so heavily that it reads as weak? Undifferentiated hedging is its own failure, and a chapter where every claim carries the same qualifier communicates nothing about which claims are solid (K3 §1). *Correct result: every claim carrying language calibrated to its evidence, with strong findings stated strongly and weak ones stated weakly.*

## 8. Analytical framework

**The interpretation ladder.** Each finding climbs this, and each rung must be earned:

    Finding (what the results showed, with its base and precision)
      → Interpretation (what it most plausibly means, with alternatives named)
        → Situation (how it stands relative to prior work: agrees, disagrees, extends, is unprecedented)
          → Implication (what follows, theoretical or practical, bounded)
            → Contribution (what is now known that was not, with its scope)

Applying it: **a claim cannot be more confident than the weakest rung beneath it** (K3 §3.6). The commonest failure is a leap from finding to contribution with the interpretation and situation rungs skipped, which produces a contribution claim with no argument behind it. The second commonest is climbing the ladder for the findings that support the study's story and not for the ones that do not.

**The disagreement diagnosis.** Applied whenever a finding conflicts with prior work, in this order:

    Population → Measurement → Method → Temporal → Analytical → Genuine

Applying it: work top down and stop at the first cause that plausibly accounts for the difference, then state it. Reaching "genuine" requires having eliminated the five above it, and a genuine disagreement stated after that work is a strong claim. A genuine disagreement asserted without the work is an unsupported one, and it is the point at which an examiner will ask whether the two studies measured the same thing.

## 9. Output format

**1. Introduction to the chapter.** What it does, and its structure.

**2. Answers to the research questions.** One short section per question, opening with the direct answer.

**3. Interpretation of each finding against the literature.**

| Finding | Agrees with | Disagrees with | No prior work | Comparability of the closest study | Diagnosed cause of any disagreement |
|---|---|---|---|---|---|

Then, in prose, the interpretation with alternatives named and confidence stated.

**4. The findings against the framework.** Per element: supported, partly, unsupported, untestable. Plus what the framework did not accommodate.

**5. Contribution.** Type or types, stated explicitly, with bounds in the sentence, and the subtraction test satisfied.

**6. Theoretical implications.** Each tied to a named finding, bounded, confidence-marked.

**7. Practical implications.** Each tied to a named finding, with the group who could act and the conditions.

**8. Unexpected and null findings.** Expectation, result, candidate explanations, most plausible reading, and for nulls the precision statement.

**9. Limitations.** Each with the affected finding and the direction of effect.

**10. Future research.** Three to five, each derived from this study, each with a design.

**11. Conclusion.** Answers the introduction's question, states the contribution, compresses the limitations, closes.

**When the study found less than hoped, the format must not force inflation (K4 §1).** A study that answered one of three questions says so, and discusses that one properly rather than padding the other two. A contribution that is contextual rather than theoretical is stated as contextual, which is honest and adequate. A null study with wide intervals reports that it could not settle the question, and its contribution is the specification of what a study that could settle it would need. Every one of these is a stronger chapter than the inflated alternative, and each is marked accordingly.

## 10. Quality checks

Run before submission. These sit on top of K4 §8.

1. Is every research question answered explicitly, in the words it was asked in?
2. Does the chapter interpret rather than restate? Would deleting the results chapter leave this one incomprehensible, as it should?
3. Is every finding situated against specific named studies rather than against "the literature" in general?
4. Is every disagreement with prior work reported, diagnosed against the six causes, and resolved or left open with reasons?
5. Has any contradicting literature been omitted?
6. Are the findings read back against the framework, including what it failed to explain?
7. Is the contribution stated explicitly, under its own heading, with its type named?
8. Does the contribution claim survive the scope, novelty and subtraction tests?
9. Does any claim extend beyond the population, setting or period the sample supports?
10. Is causal language used anywhere the design does not license?
11. Are theoretical and practical implications separated, and does each trace to a named finding?
12. Are unexpected and null findings interpreted rather than apologised for?
13. For each null, is the distinction between failing to detect and establishing absence made, and does the precision support the claim made?
14. Does every limitation name the specific finding it affects and the direction of its effect?
15. Could any future research suggestion have been written without doing this study?
16. Does the conclusion answer the question the introduction asked, in the introduction's own terms?
17. Is any strong finding hedged into invisibility, or any weak one stated with unearned confidence?

## 11. Common failure modes

| Failure | How to recognise it | How to prevent it |
|---|---|---|
| **Restatement** | The chapter walks through the results again in prose | Start from the one-sentence answers, not the results (Step 1) |
| **Unanswered questions** | The reader must assemble the answers themselves | Answer each explicitly, in its original wording (Step 1) |
| **Literature in general** | "This is consistent with the literature", no study named | Build the convergence map with named studies (Step 2) |
| **Buried disagreement** | Contradicting studies absent from a well-covered review | Report and diagnose every disagreement (Step 3) |
| **Reflexive self-blame** | Every contradiction attributed to this study's limitations | Diagnose the cause before assigning fault (Step 3) |
| **Framework abandoned** | The framework last appears in chapter three | Per-element verdict plus the residual (Step 4) |
| **Implicit contribution** | The contribution must be inferred by the reader | State it under its own heading (Step 5) |
| **Overclaiming** | A single-site study described as demonstrating a sector truth | Scope, novelty and subtraction tests (Steps 5, 11) |
| **Underselling** | A real contextual contribution described as "only" confirmatory | Contextual contributions count; state them as such (Step 5) |
| **Merged implications** | Theoretical and practical implications in one list | Separate subsections, each traced to a finding (Step 6) |
| **Asserted implications** | Recommendations that follow from no finding | Derive from named findings only (Step 6) |
| **The apologetic null** | A null result treated as a failure of the study | Interpret it, and state what the precision supports (Step 7) |
| **Generic limitations** | "Small sample, further research needed" | Name the finding affected and the direction (Step 8) |
| **Topic-level future research** | Suggestions that any study on this topic could make | Derive from this study's specific residuals (Step 9) |
| **Conclusion drift** | The conclusion answers a slightly different question | Read the introduction immediately before writing it (Step 10) |
| **Uniform hedging** | Every claim carries the same qualifier | Calibrate per claim, both directions (Step 11) |
| **AI-supplied comparison** | Literature agreement asserted for studies not read | Only studies the student has read enter the map (§12) |
| **AI-inflated contribution** | Confident novelty claims about a field's state | Novelty is a claim about a literature, and requires the literature (§12) |

## 12. AI guardrails

Skill-specific. The universal prohibitions in K4 apply in full and are not repeated. The human in the loop for every K5 marker in this skill is the candidate together with their supervisor.

1. **Academic integrity is a condition of use, not a footnote.** 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 in this skill: it will supply the convergence map structure, the disagreement diagnosis, the contribution tests, the limitation format, the overclaiming audit and a critique of the student's draft argument. It will not write the interpretation, because the interpretation is precisely what a viva examines, and a candidate who cannot reconstruct their own argument under questioning is in a worse position than one whose argument is imperfect but theirs.

2. **Never assert what the literature says about a finding without the student having read the source.** Statements such as "this is consistent with prior work showing..." are citations in waiting, and a plausible characterisation of a field is exactly the failure mode that produces a confident, wrong comparison (K4 §2.4). Ask which studies the student has read that speak to this finding, and work from those.

3. **Never supply a citation, an author, a year or a study to fill a comparison slot.** Where the student has no prior work to compare a finding against, the correct output is that no comparable work has been identified in their review, which is itself reportable and often points to a contextual contribution.

4. **Never state that a finding is novel, fills a gap, or is the first of its kind.** Novelty is a claim about the entire state of a literature, and neither the model nor a student who has read forty papers can verify it. The defensible form is bounded and search-anchored: "no study identified in this review has examined this relationship in this population". Anything stronger requires evidence nobody has.

5. **Never inflate a contribution to make the chapter feel more substantial.** A masters contribution is real and modest. Offer the contribution types and the three calibration tests, and let the student's evidence decide which claims survive. Where a draft overclaims, say so directly, since overclaiming costs more marks than the modest claim it replaced would have.

6. **Never resolve a contradiction with the literature by explaining it away.** Run the diagnosis, name the cause, and where the disagreement is genuine, leave it genuine (K4 §4.1). A discussion that dissolves every inconvenient contradiction is one an examiner will test.

7. **Never convert a null result into a positive one through language.** "Approached significance", "showed a tendency towards", "a trend was observed" applied to a non-significant result misrepresents it. Report what the estimate and its precision support, and nothing beyond.

8. **Never introduce causal language into an interpretation the design cannot support** (K4 §3.2), and never let a hedge do the work of a design. "May drive" is still a causal claim.

9. **Never generate a limitation, a future research direction or an implication that does not follow from this study's specific findings.** Generic material of this kind is easy to produce and earns nothing, and it displaces the specific material that would have earned something.

10. **Where the student's interpretation is not supported by their own results, say so plainly.** This is the most useful thing this skill can do, and the most uncomfortable. An interpretation that the results do not carry will be found in the viva, and finding it now costs a paragraph.

## 13. Best-practice principles

- **The one-sentence answer to each research question is the chapter's backbone.** Write those five sentences first and the structure follows; write them last and the chapter will have wandered.
- **Contradiction is a finding, not a failure.** The disagreement, once diagnosed, is frequently the most interesting passage in the dissertation, because it reveals that a construct is measured or behaves differently than assumed.
- **Diagnose before you attribute fault.** Your study may be better designed for this population than the one it disagrees with, and the reflex to assume otherwise costs marks.
- **A contextual contribution is a real contribution.** Examining an established relationship in a population or setting where it has not been examined is what most masters research does, and it is undersold far more often than it is overclaimed.
- **Overclaiming is more expensive than modest claiming**, because it demonstrates a failure of judgement rather than a limitation of scope, and judgement is what is being assessed.
- **Bound the claim in the same sentence as the claim.** A bound that arrives two paragraphs later has already let the overclaim out.
- **The subtraction test is the fastest contribution check available.** If nothing specific would be unknown had the study not been done, the contribution has not been found yet.
- **A limitation without a direction is a fact, not a limitation.** Say which way it would push the result.
- **Future research that could have been written before your study did no work.** Derive it from what your findings opened, not from what your topic contains.
- **The framework's failures are more informative than its successes.** What it could not accommodate is where your context exceeded the borrowed theory.
- **Read the introduction immediately before writing the conclusion.** Half of conclusion defects are questions answered slightly off-target.
- **Calibrate in both directions.** Hedging a strong finding is as much a miscommunication as inflating a weak one, and a chapter of uniform hedges tells the reader nothing.

## 14. Worked example

Generic fictional scenario, academic, in a financial services context.

**INPUT**

A masters candidate in organisational psychology studied whether flexible working arrangements are associated with reported job engagement among back-office staff in two mid-sized financial services firms in one country. Survey of 212 staff, plus 9 follow-up interviews. Three research questions: whether flexible working is associated with engagement; whether the association differs by tenure; and how staff describe the relationship between the two. The results chapter found a small positive association overall, no detectable difference by tenure, and an interview pattern in which staff attributed engagement changes less to flexibility itself than to whether their manager had visibly accepted it. The draft discussion restates the results, compares them to "the literature", claims the study "demonstrates the importance of flexible working for engagement in financial services", and lists four generic limitations.

**PROCESS**

*Step 1.* Three one-sentence answers are written. RQ1: flexible working was positively associated with engagement in this sample, with a small effect and a stated interval. RQ2: no difference by tenure was detected, with an interval wide enough that a moderate difference cannot be ruled out. RQ3: participants located the mechanism in managerial acceptance rather than in the arrangement itself.

*Step 2.* The convergence map is built. RQ1 finds broad agreement with prior work on direction but a smaller effect than most cited studies, all of which were conducted in knowledge-work settings. RQ2 has one prior study reporting a tenure interaction, in a different sector. RQ3 has almost nothing directly comparable: the literature the candidate reviewed treats flexibility as an arrangement, not as a signal.

*Step 3.* The disagreement on effect size is diagnosed. Population differs (back-office process work with limited task discretion, versus knowledge work with high discretion). Measurement differs (this study used a validated engagement scale, two of the comparison studies used single-item measures known to inflate associations). Method differs (cross-sectional in both cases, so not this). Temporal is plausible: the comparison studies predate the period in which flexible working became widespread, which changes what the arrangement signals. The diagnosis lands on population and measurement, and the discussion states that the smaller effect observed here is what would be expected in a low-discretion setting measured with a multi-item instrument, which reframes the disagreement as informative rather than as a shortfall.

*The judgement call.* The candidate's draft claim was that the study "demonstrates the importance of flexible working for engagement in financial services". Three tests are applied. Scope: two firms in one country, both mid-sized, one sector, back-office staff only, so "financial services" is population drift and "demonstrates" is a causal verb on cross-sectional data. Novelty: the association itself is well established, so the claim adds nothing where it is placed. Subtraction: if this study had not been done, what would be unknown? Not that flexibility and engagement are associated. What would be unknown is that in a low-discretion back-office setting the association is small, and that staff themselves attribute the effect to managerial acceptance rather than to the arrangement. The contribution is rewritten as contextual and empirical: the study provides evidence, in a setting where the relationship has not previously been examined, that the association is materially smaller than in knowledge-work settings, and it identifies managerial endorsement as a candidate mechanism that the existing framing does not distinguish from the arrangement itself. That claim survives all three tests, and it is a better claim than the original because it says something specific.

*Step 7.* The null on RQ2 is interpreted rather than dropped. The interval is wide and includes a tenure difference that would be practically meaningful, so the honest statement is that the study could not settle the question, not that tenure does not matter. The one prior study reporting an interaction is discussed, its sector difference noted, and the question left open with a specified design that would close it.

*Step 8.* Limitations are rewritten. The generic "small sample" is replaced by a statement that the tenure subgroup analysis is underpowered and should be read as inconclusive. Two new limitations are added that the draft lacked: the engagement measure was collected at the same time and by the same instrument as the flexibility measure, so common method variance may inflate the association, which pushes in a known direction; and both participating firms volunteered, which likely selects for organisations already positive about flexible working, biasing the association upward. Both are specific, both name the affected finding, and both give a direction.

**OUTPUT**

A discussion that opens with three direct answers, situates each finding against named studies with comparability stated, diagnoses the effect-size disagreement as a population and measurement difference rather than as a shortfall, reports the framework's failure to distinguish arrangement from managerial signal as the study's most interesting result, states a bounded contextual and empirical contribution, separates two theoretical implications from three practical ones each tied to a finding, interprets the null with its precision, lists five specific directional limitations, proposes three studies that follow from this one, and closes with a conclusion that answers the introduction's question in its own terms.

`RESEARCHER REVIEW RECOMMENDED` on the contribution statement, since whether managerial endorsement is genuinely undistinguished in the current literature is a judgement about the field's state (K5 §2.1). `RESEARCHER SIGN-OFF REQUIRED` on the practical implications, which name actions employers might take (K5 §2.3, §2.8).

## 15. Advanced usage

**Write the discussion's skeleton before analysis.** Draft the one-sentence answers for every possible result: what you would conclude if the association were strong, weak or absent. Doing this before you see the data removes the temptation to build the interpretation around whatever emerged, and it exposes in advance any result you would not know what to do with, which is a design problem worth finding early.

**The hostile reading.** Give the draft to someone with instructions to argue against every interpretation. The three attacks that arrive most often are alternative explanation, generalisation and causal direction, and pre-empting them in the text turns three viva questions into three paragraphs. See **15.19 Examiner-Proofing and Defence Strategy** for the doctoral version, and **13.04 Bias Detection** for the systematic form.

**Contribution as an abstract test.** Write the abstract's contribution sentence and check that the discussion earns it exactly, no more and no less. Abstracts are where overclaiming concentrates, because they are written last and under pressure, and they are the part of the dissertation most likely to be read by someone who reads nothing else.

**Harvesting the discussion for publication.** The contribution statement, the disagreement diagnosis and the residual are usually the publishable core of a masters dissertation, and they are often more interesting than the headline result. See **15.24 Journal Article Development**.

**Where the standard approach does not fit.** In exploratory and theory-generating studies the discussion's centre of gravity moves: the emergent conceptual account is presented here and compared to existing theory, and the contribution is the account itself rather than a finding about a relationship. In evaluative and applied work, practical implications carry more weight and must meet a higher evidential bar, since someone may act on them. In critical and interpretive traditions, the discussion may explicitly foreground the researcher's position as part of the interpretation rather than confining it to a limitation. In all cases the two non-negotiables survive: the research questions are answered explicitly, and no claim exceeds its evidence.

## 16. Skill chain

**Recommended previous skills:**
- **15.10 Data Analysis Chapter Development.** Hands over the final, completely reported findings this chapter interprets.
- **15.06 Systematic Literature Review.** Hands over the specific studies, with their populations and methods, that the convergence map is built from.
- **15.07 Theoretical and Conceptual Framework Development.** Hands over the framework the findings are read back against, including the residual.
- **15.08 Research Design and Methodology Chapter.** Hands over the design limitations this chapter extends in the light of the findings.

**Recommended next skills:**
- **15.12 Thesis Defence and Viva Preparation.** Takes the contribution statement, the diagnosed disagreements and the limitations directly into defence preparation.
- **15.24 Journal Article Development.** Takes the contribution and the disagreement diagnosis as the core of a publishable paper.
- **15.18 Thesis Architecture and Chapter Coherence**, for checking that the conclusion closes the loop the introduction opened.

**Runs well alongside:**
- **08.01 Finding to Insight Development**, for the discipline of moving from finding to explanation without skipping steps.
- **13.04 Bias Detection**, for the disconfirmation pass on the interpretation.
- **15.05 Academic Writing Structure and Argumentation**, for building the paragraph-level argument the chapter requires.

---
A Yazi Supplied Skill and resource.