Written and maintained by CASRAI Editorial Board
Last updated
A theoretical framework is only useful once it produces something a study can actually test. That happens through a specific chain: a construct (an abstract, unobservable concept) gets linked to another construct in a proposition (a general, not-yet-measurable statement of how the theory says two constructs relate), and the proposition is then translated into a hypothesis (a testable prediction stated in terms of measurable variables). This page walks through that chain step by step, with a worked example, rather than re-covering what a theoretical framework is in the abstract — for the definitional distinction between a theoretical framework (built on one existing, named theory) and a conceptual framework (a study-specific map synthesized from multiple sources), see Conceptual Framework vs. Theoretical Framework.
The reasoning chain: construct, proposition, hypothesis
These three terms sit at different levels of abstraction, and collapsing them into one another is one of the most common reasons a theoretical-framework section gets flagged in review.
| Term | What it is | Example |
|---|---|---|
| Construct | An abstract, unobservable concept the theory is built around. Not directly measurable on its own — see Research Constructs: Definition and Examples. | Perceived autonomy support; intrinsic motivation. |
| Proposition | A general, theory-level statement about how two or more constructs relate. Written in the language of the theory, not yet tied to a specific measurement instrument or population. | Perceived autonomy support increases intrinsic motivation. |
| Hypothesis | The same relationship restated as a specific, falsifiable, testable prediction about measured variables in a defined sample — see Hypothesis: Definition, Types, and How to Write a Testable One. | Employees who score higher on a validated autonomy-support scale will score significantly higher on a validated intrinsic-motivation scale than employees who score lower. |
A proposition and a hypothesis are often confused because they can describe the same underlying relationship — the difference is operationalization. A proposition stays at the construct level; a hypothesis commits to how each construct will actually be measured, in whom, and states the predicted direction in testable form.
Step 1: Anchor the framework in a named theory
A theoretical framework, as distinct from a conceptual framework built for one specific study, is built on an existing, named theory attributable to specific author(s) and applied largely intact as the study’s interpretive lens — examples include Self-Determination Theory (Deci and Ryan), the Diffusion of Innovation (Rogers), and Social Cognitive Theory (Bandura). Choosing the theory first matters because it supplies the constructs you will use and constrains what a legitimate proposition can claim: a proposition has to be something the theory itself predicts or is consistent with, not an arbitrary pairing of variables that happen to interest you.
Two practical checks before committing to a theory: first, has it actually been applied to your population or setting before, or would you be the first — either is defensible, but a reviewer will expect you to say which; second, does the theory specify a mechanism (why the relationship should hold), not just an association — a proposition without a stated mechanism reads as a hunch dressed up in theoretical language.
Step 2: Define your constructs operationally
Before a construct can appear in a proposition, it needs an operational definition — the specific, observable indicators that will stand in for the unobservable concept. This is the step covered in depth in Research Constructs: Definition and Examples (the construct-to-variable pipeline) and Construct Validity: Definition, Evidence Types, and Threats (whether your chosen indicators actually measure the construct the theory means, not something adjacent to it).
Do this before writing the proposition, not after. If you write the proposition first in loose language and only operationalize the constructs when you get to the hypothesis, it’s easy to end up with a hypothesis that doesn’t actually follow from the proposition you started with — the operational definition quietly drifts.
Step 3: Write the proposition
A proposition links two (or occasionally more) constructs in the general terms the theory itself uses. It is a claim about the theoretical world, not yet about your study’s data. Keep three things true of a well-formed proposition:
- It states a relationship between named constructs, not variables — “autonomy support” and “intrinsic motivation,” not “supervisor rating” and “self-report survey score.”
- It follows from (or is a defensible extension of) the anchor theory, not an assumption you’re smuggling in separately.
- It is directional where the theory supports a direction (“increases,” “reduces,” “moderates the relationship between”) rather than vaguely stated as “is related to,” unless the theory itself is genuinely non-directional on that point.
Worked example, continuing from Self-Determination Theory: “Perceived autonomy support from a supervisor increases an employee’s intrinsic motivation.” This is a proposition, not yet a hypothesis — it names no instrument, no sample, and no statistical test. It’s a claim the theory makes (or is consistent with) about how these two constructs relate in general.
Step 4: Derive a testable hypothesis from the proposition
Turning a proposition into a hypothesis means adding everything the proposition deliberately left out: which measured variable stands in for each construct, in which population, and what a specific test of the predicted relationship would look like. Three additions typically do this work:
- Operationalize each construct into a measured variable. Perceived autonomy support is commonly operationalized via an instrument such as the Work Climate Questionnaire; intrinsic motivation via an instrument such as the Intrinsic Motivation Inventory. The specific instrument matters — it’s what a reviewer will check construct validity against.
- Specify the population and setting. A proposition can be theory-general; a hypothesis has to name who was studied, because the predicted relationship is now a claim about a specific sample, not the construct universe.
- State the predicted relationship in testable form — direction and, where relevant, the statistical test that would support or fail to support it. See Null Hypothesis Explained for how this maps onto the null/alternative pair a statistical test actually evaluates.
Worked example, continuing the chain: “Employees who score higher on the Work Climate Questionnaire (perceived autonomy support) will score significantly higher on the Intrinsic Motivation Inventory than employees who score lower, controlling for tenure.” That is now falsifiable in a specific, checkable way the proposition was not: a study could run the correlation or regression and the prediction could fail.
Worked example, start to finish
Illustrative worked example — not a citation to a real published study. The steps below walk through the mechanics of the chain using Self-Determination Theory as the anchor; the specific numbers, sample, and result are illustrative only, constructed to show the reasoning, not reported findings.
- Theory: Self-Determination Theory (Deci and Ryan) — autonomy-supportive environments are theorized to increase intrinsic motivation by satisfying the basic psychological need for autonomy.
- Constructs: perceived autonomy support (independent) and intrinsic motivation (dependent).
- Operational definitions: perceived autonomy support = employee-rated score on a validated autonomy-support scale; intrinsic motivation = employee-rated score on a validated intrinsic-motivation inventory.
- Proposition: perceived autonomy support increases intrinsic motivation.
- Hypothesis: employees reporting higher perceived autonomy support from their direct supervisor will report significantly higher intrinsic motivation than employees reporting lower perceived autonomy support, after controlling for tenure.
- What would falsify it: no statistically significant positive relationship between the two scale scores in the sampled population, once tenure is controlled.
Notice what stayed fixed and what changed at each step: the construct pair never changed, but the language got progressively more concrete — from “autonomy support” and “intrinsic motivation” as ideas, to two named instruments, to a specific, controllable, falsifiable claim about a defined sample.
Common mistakes in this chain
- Skipping the proposition and writing a hypothesis straight from the theory. This usually produces a hypothesis that sounds theory-grounded but doesn’t actually trace back to a specific claim the theory makes — a reviewer who asks “where does the theory say this?” often exposes the gap.
- Writing the proposition using variable names instead of construct names. “Survey score increases with training hours” is already a hypothesis-shaped statement; a proposition should stay at “self-efficacy increases with mastery experience” until the operationalization step deliberately narrows it.
- A hypothesis whose constructs were never checked for construct validity. If the instrument doesn’t actually measure the construct the theory means — see Construct Validity — the hypothesis test may be statistically clean and still not test what the framework claims it tests.
- Confusing a theoretical framework’s borrowed theory with a study-specific conceptual framework. If you’re synthesizing ideas from several sources into your own model rather than applying one named theory intact, you’re building a conceptual framework, not a theoretical one — see Conceptual Framework vs. Theoretical Framework for the full distinction.
Where this goes in a manuscript, thesis, or protocol
The theoretical framework typically appears in the literature review or a dedicated “Theoretical Framework” subsection, where the anchor theory and its constructs are introduced and the proposition(s) are stated in narrative form (often alongside a framework diagram showing the constructs as boxes and the proposed relationships as arrows). The hypothesis, derived from the proposition, then moves to the end of the introduction or a dedicated hypotheses section, and is typically restated in the methods section next to the statistical analysis plan that will test it — see Hypothesis: Definition, Types, and How to Write a Testable One for that placement in more detail, and Hypothesis vs. Theory for how the finished hypothesis relates back to the theory it came from.
Frequently asked questions
What’s the difference between a proposition and a hypothesis?
A proposition is a general, theory-level statement relating constructs to each other, with no commitment to how those constructs are measured. A hypothesis restates that same relationship as a specific, falsifiable prediction about measured variables in a defined sample. Every hypothesis should trace back to a proposition; not every proposition needs to become a hypothesis in a given study.
Do I need a proposition if I already have a hypothesis?
Not formally in every write-up — many published papers move straight from theory to hypothesis in the text. But working through the proposition step explicitly, even just in your own notes, is what catches a hypothesis that doesn’t actually follow from the theory it claims to be grounded in.
Can one proposition produce more than one hypothesis?
Yes. A single proposition (“autonomy support increases intrinsic motivation”) can generate multiple hypotheses if you measure the constructs in more than one way, test the relationship in more than one population, or add a moderator or mediator to the model — each specific test is its own hypothesis, but they can share one underlying proposition.
How many constructs can a single proposition include?
Most propositions link two constructs, but a proposition can name three if the theory itself specifies a three-way relationship — for example, a moderation claim (“X increases Y more strongly when Z is high”). Beyond that, it usually becomes clearer to state it as multiple linked propositions rather than one dense sentence.
Is a theoretical framework the same as a conceptual model diagram?
The diagram is a visual representation of the framework, not a separate thing — boxes for constructs, arrows for the proposed relationships (the propositions), typically with the anchor theory named alongside it. See Conceptual Framework vs. Theoretical Framework for how this differs when the framework is conceptual rather than theoretical.
What if my data doesn’t support the hypothesis — does that disprove the theory?
Not on its own. A single unsupported hypothesis can result from measurement problems, a small or unrepresentative sample, or a genuine boundary condition on the theory, as much as from the theory being wrong. See Null Hypothesis Explained for what a non-significant result does and doesn’t establish.
