A conceptual framework is a written and drawn statement of the key variables in your study and the relationships you expect among them, built for your own research rather than lifted from a published model. Learning how to write a conceptual framework with variables is a sequence, not a design task: name the variables your question requires, break each into measurable dimensions, say in a full sentence what every arrow means, and only then draw the figure. Most frameworks that get sent back fail at one of those four points, not at the drawing.
Allow a few focused hours once your sources are in front of you. The hard part is logic, not artwork — panels are not judging whether your diagram looks good, they are checking whether your instrument can measure everything that appears in it.
Table of Contents
- 1What You Need
- 2Step-by-Step
- 31. Define the Research Problem and Main Concept
- 42. Review Theory and Existing Frameworks
- 53. Identify Concepts and Turn Them Into Variables
- 64. Define Indicators or Measurement Dimensions
- 75. Map the Proposed Relationships
- 86. Draft the Framework Diagram
- 97. Write the Narrative Explanation
- 108. Check Alignment With the Research Questions and Method
- 11Common Mistakes
- 12Frequently Asked Questions
- 13What is the difference between a conceptual framework and a theoretical framework?
- 14How many variables should a conceptual framework include?
- 15Should independent and dependent variables be shown in the diagram?
- 16Do conceptual frameworks need indicators or measurement scales?
- 17Can a conceptual framework include mediating and moderating variables?
- 18How should I cite sources in a conceptual framework diagram?
What You Need
Get these in place first. A framework built from missing sources is why so many drafts stall: there is nothing to build from, and the figure ends up as boxes with no arrows you can defend.
- A statement of the problem in one paragraph. This is the ceiling on everything that can go in the framework. A variable that never appears in the problem has no business in the figure.
- Your research questions, or your hypotheses if you already have them. Each question should map to at least one variable pair.
- Two to four theoretical or conceptual sources you can actually cite — a theory, a model, or a substantial empirical study that links the variables you care about.
- A candidate variable list. Write every term you have used loosely so far. You will cut most of them in step 3.
- Literature for each relationship you want to draw. Not literature for your topic in general, but for the specific arrow.
- A drawing tool. Word shapes, PowerPoint, Canva or draw.io all work. See step 6.
- Your department’s thesis or dissertation manual. Institutional rules beat any textbook, and format guides are where the phrase “discussed and grounded on theories, not just enumerated” comes from.
If you genuinely cannot locate a theory that fits, that is survivable. Plenty of supervisors never raise the framework at proposal stage and then flag it at the viva. Build the logic from the literature you do have and state plainly in your discussion which source supports which connection.
Step-by-Step
1. Define the Research Problem and Main Concept
Write one sentence naming the single concept your study is about, and one sentence naming the outcome you expect to explain. If you cannot do that, stop here — the rest of the framework will be a list of topics rather than an argument.
Output: a main concept plus a relationship you intend to examine. Check: a stranger could read your two sentences and guess what your study measures.
2. Review Theory and Existing Frameworks
Read the two to four sources you gathered and extract only what you need: the variables they use, the relationship each one argues, and the conditions under which it holds. Note where their unit of analysis differs from yours — a study built on individual self-efficacy cannot be bolted onto an institution-level culture measure without saying so.
Keep the theoretical framework and the conceptual framework separate. The theoretical framework is the borrowed scholarship, argued with citations. The conceptual framework is your own model of this study, built from that scholarship.
Output: a short annotated list of relationships you could defensibly borrow. Check: each item on that list has a source and a page you can point to.
3. Identify Concepts and Turn Them Into Variables
A concept is an idea; a variable is the observable, measurable form of that idea. “Employee engagement” is a concept. “Engagement, measured as the mean of six Likert items on a 1–5 scale” is a variable. The gap between the two is where weak frameworks break.
Now sort what you have left into roles and prune. The independent variable is what you think moves the outcome. The dependent variable is the outcome. A mediating variable sits on the path between them and carries part of the effect. A moderating variable changes the strength of that path. A control variable is a competing explanation you hold constant so it cannot masquerade as your result.
| Role | What it does | Arrow | Tested with |
|---|---|---|---|
| Independent | The factor you expect to drive change | Points toward the dependent variable | Correlation or multiple regression |
| Dependent | The outcome being explained | Receives the arrow | Same model, as the outcome |
| Mediating | Carries part of the effect along the path | IV to mediator, mediator to DV | Mediation analysis |
| Moderating | Strengthens or weakens the path | Points at the path itself, not at either box | Moderation analysis or interaction term |
| Control | A rival explanation held constant | Dashed line, labelled as control | Entered as a covariate |
There is no fixed number of variables. The honest answer is: as many as your questions, data and sample size can carry, and no more. One dependent variable with two or three independent variables, one moderator and one mediator is already a full model. Six moderators is a fishing expedition.
Output: a labelled variable list. Check: every entry is measurable, and every entry traces back to a research question.
4. Define Indicators or Measurement Dimensions
An abstract variable has to be observed somehow, and the route you name here is a promise to your method chapter. For a survey study that means Likert items or scales, ideally validated ones. For an interview study it means the questions you will ask. For an observational or secondary-data study it means the observable behaviours or records you will code.
Write the operational definition in plain language: the construct, the source of data, and the scale or coding scheme. Keep indicators conceptually distinct from the variable itself. Six items are indicators; the mean score is the variable.
Output: one line per variable naming its indicators. Check: walk your questionnaire item by item and confirm no box in the figure is left unmeasured.
5. Map the Proposed Relationships
Now say what every connection means, in a complete sentence, before you draw anything: “This connection means that I am proposing that…”
That frame is doing real work. It is the single strongest check on a framework, because a relationship you cannot finish that sentence about is usually an arrow you drew because the diagram looked empty. If you cannot cite literature for the connection, either find it or delete the arrow.
Be careful with direction. A single-headed arrow proposes an effect. A double-headed arrow proposes association only. Cross-sectional data can support the second and quietly undercuts the first, and panels notice when a figure claims more than the design delivers.
Output: a written relationship statement per arrow. Check: each one names a source and a testable expectation.
6. Draft the Framework Diagram
Draw the figure only now, and keep it plain: one box per variable, straight arrows, a legend explaining single-headed, double-headed, mediator and moderator arrows. Anyone should be able to read it without the full text of the paper.
For the tools, the workflow is similar everywhere. In Word use Insert > Shapes for boxes and Insert > Shapes > Lines with arrowheads for connectors, then group the whole thing. In PowerPoint use rectangles plus a line arrow from the Home ribbon’s Shapes menu, and set Align and Distribute so the layout settles into a grid. In Canva, insert rectangles, then use Elements to search for a line with an arrow end. In draw.io, add the rectangle shape, drag from one shape’s edge to another to create a connector, and set the line style to end-arrow.
For APA 7th edition, number the figure in order of appearance, place an italicised title above the image, bold or plain “Figure 1” to the left of that title, and add a note below if the model is adapted from a source.
Check: the figure is legible at print size and contains no variable that is missing from your list.
7. Write the Narrative Explanation
The figure and the discussion are two different deliverables, read by different people. The panel reads the figure in seconds; the examiner reads the argument. So write connected academic prose that explains the concepts, the theoretical basis behind each arrow, the variable definitions, your measurement choices, and your assumptions.
A useful pattern: one paragraph per arrow, opening with the relationship and closing with the source that supports it. This is the opposite of what most frameworks do. Format guides for postgraduate work keep asking for boxes that are discussed and grounded, not merely enumerated.
Check: no box in the figure is unexplained in the text.
8. Check Alignment With the Research Questions and Method
Run the alignment test in both directions. Every variable should serve a research question, every research question should be answerable with the variables you have, every variable should be measurable by your instrument, and every arrow should correspond to a named analysis in your methods chapter.
Then decide what you will do with each arrow statistically. A straight IV-to-DV arrow is a regression. A double-headed arrow is a correlation. A two-step path through a mediator is a mediation model. A variable pointing at the path is an interaction term, which is what moderation looks like in a regression table.
Check: set the figure beside your Chapter 3 outline and confirm the two match. Then remember the framework is revisable. If a pilot reveals an unmeasurable construct, redraw it — a framework approved at proposal stage is a draft, not a contract.
Common Mistakes
These are the flags that come up repeatedly in thesis panels and journal reviewers, with the fix for each.
- Arrows with no test behind them. Every arrow is a promise. If no analysis in Chapter 3 addresses it, either add the analysis or remove the arrow.
- A variable in the figure that never appears in the statement of the problem. Trace each box back. If it does not trace back, it is decoration.
- Concepts written as if they were variables. “Motivation” is not a variable until you say how you will observe and scale it.
- Vague labels. A box saying “Performance Factors” tells the examiner nothing. Name the dimension or the indicator.
- Theoretical and conceptual frameworks merged into one section. Separate them: one argues the scholarship, the other maps your study.
- An overcrowded diagram. Ten boxes with crossing arrows is unreadable. Cut to what your questions need, or split into two figures.
- Causal arrows drawn on correlational data. Use a double-headed arrow unless your design genuinely manipulates or measures change over time.
- Sources missing from the figure. Adapted models need a note beneath the figure naming the original, and your own text needs citations for every relationship you propose.
- Copying another thesis’s framework. You may reuse the structure and cite the source. You may not reuse the variables and pass them off as your model.
One more: forcing a qualitative study into IPO boxes. If your design is exploratory, a narrative framework built around sensitizing concepts, with open links and no fixed boxes, is the honest form for it.
Frequently Asked Questions
What is the difference between a conceptual framework and a theoretical framework?
A theoretical framework is the borrowed scholarship itself: the theories and prior studies you cite to justify your topic, argued in prose with references. A conceptual framework is your own constructed model of this study, naming the key variables and the relationships you presume among them, usually shown in a figure. The theoretical framework tells the reader where the ideas came from. The conceptual framework tells them what you are actually testing, and it is built, not found.
How many variables should a conceptual framework include?
There is no fixed number, and any answer that gives one is oversimplified. Include as many variables as your research questions require, your instrument can measure, and your sample size can support. A common workable shape is one dependent variable with two or three independent variables, plus a mediator or moderator where the theory calls for one. Add a variable only when it answers a question or removes a competing explanation.
Should independent and dependent variables be shown in the diagram?
Yes, and labelling them removes most panel confusion. Give the independent variable its own box with an arrow pointing to the dependent variable, then state in the text what that arrow means you are proposing. Use a double-headed arrow instead when your data is cross-sectional and supports association only. What matters more than the direction is that you say in words which relationship the arrow claims and how Chapter 3 will test it.
Do conceptual frameworks need indicators or measurement scales?
Not in the figure itself, but you need them somewhere. Indicators belong in the discussion and in the methods chapter, where you give the operational definition: the construct, the data source, and the scale or coding scheme. Keep them conceptually separate, since the six Likert items are indicators and the mean score is the variable. A framework with variables but no measurement route is the kind reviewers flag immediately.
Can a conceptual framework include mediating and moderating variables?
Yes, and they are worth including when your theory explains a path rather than a single effect. A mediating variable carries part of the effect, so you draw two arrows, from independent to mediator and from mediator to dependent, and test it with a mediation model. A moderating variable changes the strength of the path, so the arrow points at the path itself and the test is an interaction term in regression.
How should I cite sources in a conceptual framework diagram?
Give the figure an APA 7th edition number and an italicised title above it, such as Figure 1 followed by Conceptual Framework of the Study, and place the image below. If the model is adapted from a published framework, add a note beneath the figure naming the author and year. Then cite the supporting source in the discussion paragraph for each relationship, because a citation inside a diagram is not a reference a reviewer will accept on its own.
Start with the sentence, not the boxes. Write down your main concept, name the variables your research questions require, and finish the “I am proposing that…” sentence for each connection you want to draw. If those fit together and your instrument can measure them, the diagram takes about an hour. If they do not, no amount of layout work will save the framework.
