03 · What You Need to Know
A Primary Research Question Establishes What the Study Is Mainly Designed to Answer
The word “primary” should mean more than “RQ1.” A primary research question identifies the question that occupies a privileged position in the logic of the study.
In many quantitative studies, that question influences consequential design decisions. Farrugia and colleagues recommend establishing a single primary research question around which the study plan is focused. They note that additional questions can be developed, but those questions should not compromise the primary question because increasing the number of questions can complicate design, analysis, and feasibility.
This hierarchy is particularly explicit in randomized trials, where the trial is generally constructed to address its primary research question even though secondary questions may also be scientifically interesting.
Primary does not simply mean “the first question listed”
Imagine a study with these questions:
RQ1: What proportion of students use generative AI for academic writing?
RQ2: Does access to AI-supported formative feedback improve subsequent independent writing performance compared with conventional feedback?
If the sample size, intervention, randomization, primary outcome, and analysis were all designed around RQ2, then RQ2 is functionally the primary research question even if it appears second on the page.
Numbering cannot establish methodological priority.
Primary research question
The central question the study is principally designed and resourced to answer.
Secondary research question
An additional prespecified question that contributes useful information but does not replace the study's principal inferential target.
The primary question often determines the primary outcome
In intervention and other confirmatory quantitative studies, the primary research question and primary outcome are closely connected.
Suppose the question is:
“Among first-year university students, does retrieval-practice instruction compared with rereading improve delayed recall one week later?”
The primary outcome needs to represent the answer to that question. If delayed recall is the intended outcome, the study should not quietly become a study of immediate quiz performance merely because that result happens to be more favorable.
Clinical-trial methodology similarly emphasizes determining the primary question before defining the primary outcome used to answer it. Secondary outcomes can provide additional information, but the primary outcome retains a special role in evaluating the principal question.
The primary question can drive sample-size planning
Sample size cannot sensibly be planned in the abstract. In many quantitative studies, it depends on the primary outcome, expected or important effect, desired precision or power, variability, event rate, design, and planned analysis.
If you have several equally “primary” questions, they may require different sample sizes.
A study adequately powered to detect a difference in examination scores might be underpowered to detect a rare adverse outcome or an interaction between subgroups. This is one reason methodological guidance recommends focusing resources around one primary question rather than assuming a single sample will automatically answer every interesting question.
Secondary analyses can still be informative, but the evidentiary strength of their conclusions should reflect what the study was actually designed to support.
A primary question helps choose the study design
Research design should fit the question. A question about prevalence may call for a very different design from a question about causal effects, lived experience, diagnostic accuracy, or prediction.
If a study has several co-equal questions requiring fundamentally different designs, the problem may not be that multiple questions are prohibited. The problem may be that the project lacks one coherent design.
Research-methods guidance consistently treats the research question as foundational to study design. A clearly articulated question helps determine what evidence needs to be collected and which design is capable of producing it.
This is also why two questions requiring different methods need deliberate methodological integration rather than simply being placed in the same proposal.
Secondary questions can add value without competing with the primary question
A secondary research question is not an unimportant question. It simply occupies a different position in the study's hierarchy.
Suppose the primary question asks whether an educational intervention improves delayed learning performance. Secondary questions might examine student engagement, intervention adherence, another prespecified learning outcome, or whether implementation differed across sites.
Those questions can enrich interpretation of the primary finding.
The important constraint is that secondary questions should not consume so much sample, measurement burden, analytical complexity, or attention that the study can no longer answer the primary question adequately.
Methodological guidance explicitly recognizes primary and secondary aims and objectives while emphasizing that they should remain closely connected to the research question.
Secondary does not mean post hoc
A secondary question can be planned before the study begins.
For example:
Primary question: “Does the intervention improve writing performance?”
Secondary question: “Does the intervention improve writing self-efficacy?”
If both are established during protocol development, the second is a prespecified secondary question.
Now suppose the researcher examines 25 questionnaire variables after data collection, discovers that one differs significantly between groups, and creates a new question about it. That is different. The analysis may still be worth reporting, but it should be identified as exploratory or post hoc rather than retrospectively presented as a prespecified secondary question.
Methodological guidance emphasizes defining research questions and hypotheses before the study, partly because working backward from statistically interesting findings can increase the risk of spurious conclusions.
A primary question helps distinguish confirmatory from exploratory work
In confirmatory research, investigators typically seek strong evidence concerning a prespecified claim. The primary question identifies the claim the study was principally designed to evaluate.
Exploratory analyses serve another purpose. They may reveal patterns, generate hypotheses, identify potential mechanisms, or suggest questions for subsequent research.
Both are scientifically useful, but their evidentiary roles differ.
When a study reports dozens of analyses without distinguishing the central question from exploratory findings, readers may have difficulty determining which conclusions were planned and which emerged from the data.
A primary-question hierarchy can make that distinction more transparent.
One primary question can reduce problems created by multiplicity
When a study tests many hypotheses, the probability of obtaining apparently noteworthy findings by chance can increase. The precise implications depend on the inferential framework, relationships among outcomes, statistical procedures, and whether analyses are confirmatory or exploratory.
Defining one primary question does not magically eliminate multiplicity. Nor does every secondary analysis require the same adjustment procedure.
What a hierarchy does provide is clarity about which inference the study was chiefly designed to support.
This is particularly useful when multiple research questions begin to create substantial analytical burden.
The primary question should not be selected after seeing the results
Suppose a study measures three outcomes: examination performance, self-efficacy, and engagement. The protocol does not specify which is primary.
After analysis, examination performance shows little difference, self-efficacy changes modestly, and engagement produces a statistically significant result. The researcher then labels engagement as the “primary outcome.”
That reverses the intended logic of prespecification.
For confirmatory research, the primary question and corresponding outcome should be established before the results are known. Otherwise, the label “primary” becomes a description of which finding looked best rather than which question drove the study.
Watch Out
Do not promote a secondary or exploratory question to primary status because it produced the most favorable result. In confirmatory research, primary status should reflect the study's prespecified design and inferential priorities, not the eventual P value.
Not every quantitative study needs a primary hypothesis
The distinction between a primary question and a primary hypothesis is important.
A descriptive study might have the primary question:
“What proportion of undergraduate students use generative AI for assessed academic writing?”
That can be the study's primary question even if there is no directional hypothesis. The study may be designed to estimate a prevalence with useful precision rather than test whether it exceeds an arbitrary threshold.
Descriptive research therefore can have a clear primary question without manufacturing a hypothesis merely to fit a confirmatory template. As discussed earlier, a research question can be descriptive without asking about relationships or effects.
Qualitative studies may use a central question instead
Qualitative research often follows a different organizational logic.
A study might ask:
“How do first-year university students navigate expectations concerning generative AI in academic writing?”
Supporting questions could explore how students interpret policy, negotiate differences among instructors, or decide when AI assistance is acceptable.
Calling the overarching question “primary” would not necessarily be wrong, but “central research question” may better reflect the methodological structure. The supporting questions function as subquestions that help explore dimensions of the central phenomenon rather than as secondary outcomes competing for statistical priority.
This difference follows from the broader distinction between qualitative and quantitative research questions. The structure should fit the methodology rather than importing clinical-trial terminology into every research tradition.
Mixed-methods studies may have more than one kind of priority
Mixed-methods research complicates the simple primary-secondary model further.
A study might have:
Quantitative question: “How frequently do students use generative AI for academic writing?”
Qualitative question: “How do students explain their decisions about when to use generative AI?”
Integration question: “How do students' explanations help interpret the quantitative patterns of generative AI use?”
Depending on the design, one component may have priority, or qualitative and quantitative strands may be given comparable emphasis.
Forcing one question to be “primary” merely to satisfy a generic rule could obscure rather than clarify the design. What matters is that the relationship among the questions and the intended integration are explicit.
A dissertation may contain several substantial questions
Dissertations and other large projects often have greater scope than a single empirical article. A dissertation may contain several phases, datasets, or linked studies.
In such cases, one overarching question or aim may connect several substantial research questions. Alternatively, each study within a manuscript-based dissertation may have its own primary question.
The appropriate structure depends on institutional requirements, disciplinary conventions, and the dissertation design. Researchers should therefore verify local requirements rather than treating guidance developed for clinical trials as universal dissertation law.
“Primary” should reflect intellectual priority, not merely administrative convenience
A question should not be declared primary solely because a form requires something in the “Primary Objective” field.
The designation should answer a substantive question: If this study could answer only one question convincingly, which answer would justify having conducted the study?
That thought experiment can be surprisingly revealing.
If the answer is obvious, you probably already have a primary question even if you have not labeled it. If every question seems equally indispensable, examine whether the study truly has several co-equal aims or whether its scope has become too large.
The primary question should connect directly to the study's purpose
Research questions, aims, objectives, hypotheses, outcomes, and analyses are related but not interchangeable.
| Element |
What it does |
Example |
| Research problem |
Identifies the uncertainty or issue motivating the study |
Little is known about whether AI-supported feedback improves later independent writing |
| Primary research question |
States what the study principally seeks to answer |
Does AI-supported formative feedback improve subsequent independent writing performance compared with conventional feedback? |
| Primary objective |
States what the study will do to answer the primary question |
To compare subsequent independent writing performance between students receiving AI-supported and conventional feedback |
| Primary outcome |
Defines the principal measured result used to address the question |
Score on a prespecified independently completed writing assessment |
| Secondary question |
Addresses an additional prespecified uncertainty |
Does AI-supported feedback change writing self-efficacy? |
| Exploratory question |
Investigates additional patterns or hypotheses without the same confirmatory status |
Do observed intervention differences appear to vary according to prior AI experience? |
Keeping these concepts distinct prevents the proposal from becoming a collection of labels that all appear to mean “what we are studying.”
Do not create a primary question merely by combining everything
Suppose your study asks about AI-use prevalence, writing performance, academic integrity, student experiences, faculty perceptions, and institutional policy.
Writing an overarching question such as “How does generative AI affect higher education?” does not solve the scope problem. It simply places a broad umbrella over several different inquiries.
A primary or central question should organize the study, not conceal its complexity.
If the supposed primary question is so broad that its secondary questions require largely independent studies, revisit whether the research question itself has become too broad.
Secondary questions should be interpretable even when the primary result is null
Suppose an intervention shows no convincing difference on the primary outcome. Can you still examine prespecified secondary outcomes?
Often yes, but their interpretation depends on the design, analysis plan, multiplicity strategy, and inferential framework. A null primary finding does not automatically erase all secondary evidence, nor does a positive secondary finding automatically rescue the study's primary hypothesis.
This distinction is another reason to establish the hierarchy before analysis. Readers can then interpret each result according to the role it was intended to play rather than treating whichever result is most interesting as the study's new centerpiece.
The hierarchy should be established during study planning
If a primary-secondary structure is appropriate, establish it before data collection and, for confirmatory research, before examining outcome results.
Farrugia and colleagues explicitly recommend developing primary and secondary questions during the planning stages and caution that additional questions should not compromise the primary question. They also emphasize that the primary research question forms the basis of the study hypothesis and objectives.
Prespecification does not prohibit intellectual flexibility. Unexpected findings can generate valuable new questions. The distinction is that later questions should be identified honestly as later questions.
The best structure is the one that accurately represents the study
A useful hierarchy might be:
Primary question → secondary questions → exploratory analyses.
Another study might use:
Central question → subquestions.
A mixed-methods study might use:
Quantitative question + qualitative question → integration question.
The choice should follow the study's methodological logic.
The central requirement is transparency. Readers should understand what the study was principally designed to answer, what additional questions were prespecified, what emerged later, and how much evidentiary weight each answer can reasonably carry.
