A conceptual framework is the model you build for your own study: the constructs it uses, how each is defined, the relationships you expect between them, and the conditions under which you expect those relationships to hold. A theoretical framework is an existing theory you adopt. Most postgraduate theses need both, and examiners penalise candidates who cannot say which is which.
What follows is the framework as a doctoral or Master’s candidate has to build it: its four compulsory parts, five worked examples showing what those parts contain in different disciplines, the diagram conventions examiners read, and how the framework has to reappear in the analysis. Which existing theory an education doctorate should adopt is a separate question, answered in our piece on theoretical frameworks for an education PhD or EdD. Research degree regulations differ by institution; check where your own code of practice expects the framework to sit.
How is a conceptual framework different from a theoretical one?
By authorship and by scope. A theoretical framework is someone else’s: Bourdieu’s, Engeström’s, an established model in your field, cited from its canonical source and applied to your question. A conceptual framework is yours: it may borrow constructs from one or several theories, add constructs the theories do not supply, and specify relationships nobody has yet proposed. The theory explains a class of phenomena; your framework explains the particular one you are studying.
The practical test is whether the framework could be wrong. A theoretical framework you have merely adopted cannot be falsified by your study, because you are using it, not testing it. A conceptual framework makes commitments — these constructs, this direction of effect, these boundaries — and your findings can contradict them. If nothing in your framework could be contradicted by your data, you have restated a theory rather than built a framework.
What four parts must it contain?
Every defensible framework has the same four components, whatever the discipline. Missing any one of them is the commonest structural fault an examiner will name.
| Part | What it is | How it fails |
|---|---|---|
| Constructs | The named things your study is about, typically three to six | Too many, or named so broadly that two of them overlap |
| Definitions | What each construct means in this study, sourced where a source exists | Dictionary definitions, or a definition that drifts between chapters |
| Propositions | The expected relationships, stated as sentences and numbered | Arrows on a diagram with no sentence behind them |
| Boundary conditions | The population, setting and period within which you expect it to hold | Omitted, so the framework implicitly claims universality |
Boundary conditions are the part candidates skip and the part that protects them. A framework with stated boundaries cannot be attacked for failing outside them; a framework without them invites the viva question about why your findings did not replicate somewhere you never claimed to be describing.
What do conceptual frameworks look like in practice?
Five worked outlines follow, each set out in the four parts. They are illustrative specimens — the constructs are plausible for the discipline, but the propositions are examples of the form, not findings. Replace each with constructs you can define from your own literature.
Management and the DBA
Constructs: perceived supervisory monitoring; manager trust in the team; co-presence; retention intention. Definitions: each operationalised from an existing instrument where one exists, from the literature where it does not. Propositions: P1, reduced co-presence increases perceived monitoring; P2, perceived monitoring reduces retention intention; P3, manager trust weakens the P1 relationship. Boundaries: knowledge-work teams in organisations with permanent hybrid policies, in one national employment-law context, observed after the policy became permanent rather than during transition.
Note that P3 is a moderating claim, which the diagram has to show differently from P1 and P2. How a management doctorate converts propositions like these into testable hypotheses and operationalised variables is set out in our guide to research questions, hypotheses and variables for a management doctorate; a professional doctorate carries the same requirement with a practice edge, as our comparison of the professional doctorate and the PhD explains.
Education and the EdD
Constructs: curricular intent; enacted classroom practice; teacher subject confidence; pupil access to the material. Definitions: intent from policy documents, enactment from observation, confidence from a self-report measure, access defined in terms of the specific curricular content. Propositions: P1, the gap between intent and enactment widens where subject confidence is lower; P2, that gap reduces pupil access; P3, departmental collaboration narrows the gap independently of confidence. Boundaries: one key stage, one subject, secondary settings in a single jurisdiction, within one curriculum cycle.
This is the case where the conceptual framework sits on top of a theoretical one: Bernstein or activity theory supplies the account of how knowledge is recontextualised, and the framework above specifies what you will actually measure within it.
Health services research
Constructs: protocol fidelity; routine compatibility; staffing adequacy; patient-facing outcome. Definitions: fidelity against the published protocol, compatibility as the degree of fit with pre-existing ward routines, adequacy from establishment data. Propositions: P1, routine compatibility predicts fidelity; P2, staffing adequacy predicts fidelity independently; P3, fidelity mediates the relationship between both and the outcome. Boundaries: one intervention, one care setting type, within the governance regime of a single health system.
P3 is a mediating claim and is the whole point of the framework: it says the two candidate explanations do not act directly on the outcome but through delivery. That is a commitment your data can refute, which is what makes it a framework.
Information systems and engineering management
Constructs: tool capability; workflow disruption at adoption; user discretion; sustained use. Definitions: capability from documented feature sets, disruption as measured change in task sequence, discretion as whether use is mandated. Propositions: P1, disruption at adoption reduces sustained use; P2, capability offsets disruption only above a threshold; P3, where use is mandated, discretion no longer moderates. Boundaries: professional users in organisations of a stated size band, over a defined post-adoption window.
P2 states a non-linear relationship, and frameworks that contain one must say so explicitly, because a reader will otherwise assume every arrow is linear and monotonic.
Criminology and social policy
Constructs: frontline discretion; procedural workload; organisational signal about priorities; case outcome variation. Definitions: discretion as the decision space formally available, workload as cases per practitioner, signal from internal guidance and performance measures. Propositions: P1, procedural workload narrows effective discretion below its formal level; P2, organisational signals redirect rather than reduce discretion; P3, the combination produces systematic rather than random outcome variation. Boundaries: one agency, one decision point, one policy period.
How should the diagram be drawn?
Examiners read framework diagrams as claims, so the conventions matter more than the graphic design. Boxes are constructs and nothing else — do not put methods, data sources or chapter numbers in boxes. A solid arrow is a hypothesised direct relationship. An arrow that lands on another arrow is a moderator. A construct sitting in the path between two others is a mediator, and must be drawn in the path rather than floating beside it. A dotted outline marks the boundary of the study, with anything outside it shown as context rather than as something you measure.
Then the rule that saves the most viva time: number every arrow with its proposition, and make sure every numbered proposition appears as a sentence in the text. A diagram with eight arrows and four propositions tells the examiner that four of your claims have never been written down, and that is exactly where the questioning starts.
Where does it sit, and how does it drive the analysis?
Usually at the end of the literature review or as a short chapter of its own immediately after it, because the framework is the output of the synthesis rather than an input to it — our guide to building the literature review chapter covers the synthesis that produces it. It should follow directly from the gap and problem you have already stated; if your constructs do not appear in your gap and problem statements, one of the two is wrong.
Its real work happens later. The analytic categories in your findings chapters must be the framework’s constructs, under the same names. The discussion must return to each proposition and say whether it was supported, unsupported or not testable with the data obtained. And the conclusion should state what the framework could not explain, because a framework that explained everything was either trivial or was not tested — which is the same judgement examiners apply to an original contribution to knowledge.
What does a Master’s framework do differently?
It is smaller and it may be adopted rather than built. A Master’s dissertation can legitimately take a published framework, state that it is doing so, and adapt one or two constructs for the setting — declaring the adaptation and justifying it is the whole requirement. Three constructs and two propositions is a reasonable scale, and boundary conditions still have to be stated.
What a Master’s candidate should not do is build a framework with six constructs and then test one relationship. The framework has to be answerable by the study attached to it; an oversized framework is read as an overclaim, not as ambition.
Which failures do examiners name?
Five recur. A diagram with no propositions written out underneath it. Constructs that overlap, so two boxes are measuring the same thing. Definitions taken from a textbook summary rather than from a source, and then used inconsistently by chapter six. No boundary conditions, so the framework silently claims to apply everywhere. And the structural one, which is visible from the contents page: a framework introduced in chapter three that never reappears until the conclusion. If your findings chapters use headings that are not your constructs, the framework has stopped doing work, and the viva will establish that within ten minutes.
Keeping the constructs stable for four years
Framework drift is slow and hard to see: a construct defined one way in year one is used slightly differently in the findings written in year three, and the discussion quietly reconciles the two. The defence is to keep the definitions in one place and write the findings against them rather than from memory. Draft your thesis in Tesify and hold the constructs, the numbered propositions and every chapter in one workspace, so that the framework you defend is the one your analysis actually used. The model and the argument stay yours; there is a free plan, and the useful first step is to write your three to six constructs with one definition each.
Frequently asked questions
Do I need both a theoretical and a conceptual framework?
Most postgraduate theses have both and should say which is which. The theoretical framework is the existing theory you adopt; the conceptual framework is the model you build for your study, which may draw constructs from that theory and add others. A thesis that presents only a theory has not yet specified what it will examine.
How many constructs should a conceptual framework have?
Three to six for a doctoral study, three or four for a Master’s dissertation. Beyond that, constructs start to overlap and no single study can test the relationships between them. If you have eight, two pairs are usually the same thing under different names.
Can I adapt someone else’s framework?
Yes, and it is often the stronger move, provided you say explicitly what you changed and why. An adapted framework with a justified modification is defensible; an adopted framework presented as though you built it is not, and an examiner who knows the original will notice.
Does a qualitative thesis need a conceptual framework?
Yes, though it works differently. In qualitative work the framework is often provisional, sensitising the analysis rather than specifying testable relationships, and it may be revised as the fieldwork proceeds. Say that it is provisional, and show in the discussion how the data changed it.
Where should the framework diagram appear?
At the end of the chapter that develops it, after the constructs and propositions have been set out in prose, not before. A diagram that arrives first is read as decoration; one that arrives last is read as a summary of arguments already made.
What is the difference between a proposition and a hypothesis?
A proposition states an expected relationship between constructs in conceptual terms. A hypothesis states it between measured variables, in a form a statistical test can evaluate. Quantitative theses usually need both, with each hypothesis traceable to a numbered proposition.
Should the framework change during the PhD?
It often does, and that is not a failure. What examiners object to is a framework chapter describing the original model while the findings chapters analyse a different one. Revise the chapter once the analysis is stable, and say in the discussion what the data changed.
How do I define a construct properly?
Give the definition you are using, attribute it to the source you took it from, and state any modification you made for your setting. Then use exactly that wording every time the construct appears. A construct defined from a textbook summary rather than a primary source is the version examiners query first.
What are boundary conditions and do I have to state them?
They are the population, setting and period within which you expect the framework to hold. State them: a framework without boundaries implicitly claims to apply universally, which no postgraduate study can support, and stating them protects you from questions about cases you never claimed to describe.