Research Question Development: Scope, Answerability and Design Fit

A research question is not just a polished sentence placed near the beginning of a proposal. It is a decision tool that defines what the study is trying to understand, what evidence would be relevant, which methodological choices can be justified, and whether the project is realistic within the available time, access and resources.

Strong research questions are usually developed iteratively. A student may begin with a broad topic, examine the literature, identify a meaningful problem or uncertainty, test what evidence can realistically be obtained, discuss the idea with a supervisor, and then revise the question several times. The wording improves because the underlying research logic improves.

Reader outcome

After using this guide, you should be able to move from a broad topic to one or more research questions that are clearer, more focused and more answerable. You should also be able to test whether the questions align with the literature, the available evidence and a feasible research design.

This guide does not prescribe one universal wording formula. Different disciplines express research questions differently, and some projects use hypotheses, propositions, objectives or design briefs alongside or instead of conventional question wording. Your programme requirements and disciplinary conventions remain controlling.

Start with the research problem, not the sentence

Students often spend too much time trying to make the wording sound academic before they have decided what the research problem actually is.

A better starting sequence is:

broad topic → research context → specific problem or uncertainty → relevant literature → possible contribution → research question

For example, consider the broad topic:

Social media and university students.

This is not yet a research problem. It does not tell us which social media activity matters, which students are relevant, what outcome or experience is being investigated, or why the study is needed.

A more developed problem might be:

Postgraduate students increasingly use short-form social media for academic information discovery, but the literature in a particular programme context does not yet clearly explain how students judge the credibility of research-related content encountered through those platforms.

That problem can support a more focused question because the important decisions are becoming visible: population, context, phenomenon and intellectual purpose.

The central lesson is that a research question becomes easier to write when the underlying problem is already clear.

Treat question development as an iterative process

Monash University guidance emphasizes that researchers generate and refine specific research questions from a topic and an initial understanding of the literature. This is useful because it challenges the idea that the first question you write must remain unchanged.

A practical refinement cycle can look like this:

1. Write the broad topic in one sentence.
2. Identify the specific problem or uncertainty that interests you.
3. Read enough current literature to understand how the field defines and studies the issue.
4. Draft a provisional question.
5. Test what evidence would be needed to answer it.
6. Test whether that evidence is realistically accessible.
7. Check whether the question fits the available time and resources.
8. Check whether the likely research design can address the question.
9. Discuss the question with a supervisor where appropriate.
10. Revise and repeat until the question and project logic align.

Revision does not necessarily mean the original idea was poor. It often means the project is becoming more researchable.

Use five tests for a workable research question

A useful research question should pass five connected tests:

1. Clarity: Can a reader understand what is being investigated?
2. Scope: Is the question narrow enough for the project but substantial enough to justify research?
3. Literature fit: Does the question emerge from a defensible understanding of existing scholarship?
4. Answerability: Can relevant evidence actually be obtained and analysed?
5. Design fit: Can a justified research design address the question within ethical and practical constraints?

These tests are more useful than asking whether the wording simply sounds academic.

Test 1: Clarity

A clear question identifies the intellectual task without requiring the reader to guess what the main concepts mean.

Compare:

How does technology affect students?

with:

How do first-year postgraduate students in [context] describe the role of generative AI in planning and revising assessed academic writing?

The second question is clearer because it specifies a population, a phenomenon and the type of understanding being sought. It is still only an example. It does not prove that this is the best population, that interviews are appropriate, or that the study is original.

Check vague terms

Words such as these often require clarification:

• impact;
• effect;
• influence;
• success;
• performance;
• experience;
• quality;
• engagement;
• effectiveness;
• behaviour;
• perception.

The words themselves are not wrong. The problem is that they can hide several different research meanings.

If your question asks about “impact,” ask what evidence would demonstrate that impact. If it asks about “experience,” ask whose experience, in what context and concerning which aspect of the phenomenon.

Test 2: Scope

A question can be clear but still be too large for the project.

Broad scope often appears through several dimensions at once:

• too many populations;
• too many countries or settings;
• too many variables or concepts;
• several unrelated outcomes;
• multiple methodological purposes;
• a very long historical period;
• a question that effectively contains several separate studies.

Consider:

What is the impact of artificial intelligence on education, employment and society?

This may be a legitimate broad area of inquiry, but it is unlikely to be a workable dissertation question without major narrowing.

A scope audit can help:

Narrowing is not the same as making a study trivial. A well-bounded question often allows deeper analysis because the project can devote more attention to the most important relationships or mechanisms.

Test 3: Literature fit

King’s College London and the University of Sussex both connect research questions to existing scholarship in their proposal guidance. This reflects an important principle: a question should be situated in what is already known, debated or uncertain.

The literature helps you determine:

• whether the terminology in your question matches the field;
• whether the problem has already been studied extensively;
• whether important theoretical perspectives are relevant;
• whether findings conflict across contexts or methods;
• whether your proposed contribution is plausible;
• whether the question is based on an assumption that the literature does not support.

Do not manufacture originality

A research question does not become original merely because it includes a new country, organisation or population.

Likewise, a low number of search results does not prove a research gap.

A more defensible process is:

1. Review the relevant literature.
2. Identify what the evidence currently establishes.
3. Identify important uncertainty, disagreement, limitation or contextual concentration.
4. Decide whether your proposed question genuinely addresses that issue.
5. Phrase the rationale according to the strength of the evidence.

If originality is still uncertain, use cautious language rather than making an absolute claim.

For example:

Existing evidence is concentrated in large public universities, creating uncertainty about whether the same pattern applies in smaller specialist institutions.

This is more defensible than:

No research has examined this topic in smaller institutions.

unless you have evidence strong enough to support the absence claim.

Test 4: Answerability

A research question can be interesting and literature-informed but still fail because the evidence needed to answer it cannot realistically be obtained.

Ask:

• What evidence would count as an answer?
• Where does that evidence exist?
• Can I legally and ethically access it?
• Can participants realistically be recruited?
• Is the dataset available at the required level of detail?
• Can the necessary documents, archives, organisations or technical systems be accessed?
• Are there language, cost, software or equipment barriers?
• Can the evidence be obtained within the project timeline?

Example: inaccessible population

Suppose the question is:

How do chief executives of multinational pharmaceutical companies make strategic decisions about AI investment?

The question may be intellectually interesting, but a student without a credible access route to those executives may have an answerability problem.

Possible redesigns could include:

• studying publicly available strategic disclosures;
• focusing on managers who are realistically accessible;
• selecting a bounded case-study context;
• reframing the question around organisational documents rather than elite interviews.

The correct change depends on the research purpose. The key point is that access should influence question design before the project becomes locked in.

Test 5: Design fit

Research design should follow the intellectual purpose of the question. It should not be selected mechanically from one keyword.

For example, the word “relationship” does not automatically require one particular statistical method. The word “experience” does not automatically mean interviews. The word “effect” does not prove that a causal design is feasible.

King’s College London and Sussex both connect methodology to the central research question in their proposal guidance. The underlying principle is alignment.

Ask:

• What exactly would constitute evidence for the answer?
• Is the question descriptive, interpretive, comparative, explanatory, predictive, evaluative, causal, design-oriented or something else?
• What epistemological assumptions matter in the discipline?
• What data or material can represent the phenomenon adequately?
• What analysis could transform that evidence into a defensible answer?
• Which limitations would remain even after the analysis?

A question-method alignment table

This table is not a method-selection algorithm. It is a reminder that the intellectual purpose of the question constrains what kind of evidence and analysis can be justified.

Distinguish topic, problem, aim, objective and question

These elements are related but not interchangeable.

Topic

The broad area of interest.

Generative AI in postgraduate education.

Research problem

The specific issue, uncertainty or limitation that makes investigation necessary.

Students increasingly use generative AI for academic writing, but the way they evaluate and revise AI-generated suggestions in a particular learning context is not yet well understood.

Aim

The overall purpose of the study.

To examine how postgraduate students evaluate and revise generative-AI suggestions during academic writing.

Research question

The inquiry the study will answer.

How do postgraduate students in [context] evaluate and revise generative-AI suggestions during assessed academic-writing tasks?

Objectives

The practical research tasks that help accomplish the aim.

Possible objectives might include identifying common evaluation criteria, examining revision decisions and comparing patterns across relevant student characteristics, if those objectives genuinely fit the design.

The exact way your programme distinguishes aims, objectives and questions may differ. Use the required terminology rather than forcing this example onto another discipline.

Avoid combining several studies into one question

A common problem is the overloaded question:

How does remote work affect productivity, mental health, leadership, communication, innovation and organisational culture among employees and managers across different industries?

This question contains many constructs, levels of analysis and possible outcomes.

You can often detect overload by counting the major decisions hidden inside the sentence.

Ask:

• How many phenomena are being investigated?
• How many outcomes?
• How many populations?
• How many contexts?
• How many distinct comparisons?
• Would answering one part still leave several independent studies unfinished?

If the answer is yes, consider one central question with a small number of tightly connected subquestions.

Use subquestions only when they support the central question

Subquestions can help divide a complex inquiry, but they should not become unrelated mini-projects.

A good subquestion should:

• contribute directly to answering the main question;
• use compatible evidence and design logic;
• fit within the same project scope;
• avoid duplicating the main question in different wording.

For example, a central question might ask how international master’s students adapt their academic-writing practices after beginning to use generative AI.

Possible subquestions could focus on:

• which stages of writing they use the tools for;
• how they decide whether to accept or reject suggestions;
• what concerns influence their use decisions.

These could form one coherent qualitative inquiry if the design supports them.

A weak set of subquestions might add unrelated questions about university policy effectiveness, labour-market outcomes and national AI regulation. Those issues may be interesting, but they could require different evidence and separate projects.

Check whether the question assumes its own answer

Some questions contain built-in conclusions.

For example:

Why does social media reduce students’ academic performance?

This assumes that social media reduces performance before the study has established that relationship.

A more neutral version could be:

What relationship exists between [defined social-media behaviour] and [defined academic-performance measure] among [population]?

or, depending on the research purpose:

How do students describe the ways social-media use interacts with their study practices?

The correct version depends on the evidence and design. Neutral wording does not mean avoiding theory or hypotheses. It means not pretending the empirical result is already known when it is actually the subject of investigation.

Check feasibility before finalizing wording

Sussex guidance connects research questions with practical methodology, participant or data selection, ethical considerations and timeline. This means feasibility is not something to check after the question has been approved. It belongs inside question development.

Use a feasibility matrix:

If a major area is red, revise the research question before investing heavily in proposal writing or data collection.

Use supervisor feedback as a refinement tool

A useful supervisor conversation is more specific than asking:

Is my question good?

Instead, bring a short reasoning note containing:

• the broad topic;
• the specific problem;
• the provisional question;
• what the literature currently suggests;
• what evidence you think is required;
• the proposed design;
• the main feasibility risk;
• the exact decision on which you need feedback.

Questions for a supervisor might include:

• Is the scope realistic for this degree level and timeline?
• Does the question follow logically from the stated research problem?
• Am I assuming access to evidence that may be unrealistic?
• Is the question too descriptive for the intended contribution?
• Does the proposed design actually allow the question to be answered?
• Which disciplinary convention am I overlooking?

Question refinement should remain open to feedback because literature, access and methodological feasibility can change during planning. Revision is normal. It does not guarantee supervisor or programme approval.

A worked refinement example

Suppose a student begins with:

The impact of remote work on employees.

Step 1: identify the problem

The student notices that much of the discussion treats remote work as a single condition, while employees may experience very different levels of autonomy, communication demand and managerial monitoring.

Step 2: identify the context

The student has realistic access to professional-services employees in one organisation that uses a hybrid working model.

Step 3: identify the intellectual purpose

The student is interested in how employees interpret autonomy and monitoring within hybrid work, rather than estimating a population-level causal effect.

Step 4: draft a question

How do professional-services employees in [organisation/context] describe the relationship between perceived autonomy and managerial monitoring in hybrid work?

Step 5: test answerability

The student can realistically recruit employees, but access to confidential productivity metrics is unlikely.

Step 6: test design fit

If the purpose is to understand employee interpretations, an appropriately justified qualitative design may be more aligned than a causal-effect design. That conclusion must still be justified through the discipline and methodology literature.

Step 7: revise if necessary

If ethics, organisational access or recruitment becomes difficult, the population or evidence source may need to change.

The example shows why question development is not mainly a wording exercise. Each revision follows from research logic and feasibility.

Common research-question failures

Starting with a method instead of a problem

I want to use SEM, so what question can I ask?

A method can be part of your skill set, but it should not determine the research problem. Start with the question and evidence needs, then justify the method.

Treating keywords as method instructions

A question containing “effect” does not automatically prove causal identification. A question containing “relationship” does not automatically validate a particular correlation model. A question containing “experience” does not automatically justify interviews.

Making the question impressive by making it broad

A broad question can hide weak planning. Research quality comes from defensible alignment, not from how many concepts appear in the title.

Adding a country name and calling it a gap

A new context may matter, but contextual novelty must be theoretically or practically meaningful. Location alone does not prove a contribution.

Writing the conclusion into the question

Avoid assuming the direction, existence or importance of a relationship unless the question is explicitly framed to test a theory or hypothesis and the design can support that test.

Ignoring access until after approval

A question that depends on inaccessible participants, confidential records or unavailable equipment is not ready merely because the wording is clear.

Freezing the question too early

Early questions are provisional. Literature review, pilot work, ethics requirements, access problems and supervisor feedback can all justify revision.

A practical research-question development workflow

Stage 1: Write the broad topic

Use one sentence. Do not worry about academic wording yet.

Stage 2: State the specific problem

What is uncertain, contested, unexplained, poorly understood, practically unresolved or methodologically limited?

Stage 3: Check the literature

Identify how the field defines the issue, what has already been studied and what claims can be made defensibly.

Stage 4: Draft the provisional question

Write the simplest sentence that captures the inquiry.

Stage 5: Define the evidence needed

What data, participants, texts, cases, measurements, observations or technical outputs would count as evidence?

Stage 6: Test feasibility

Check access, ethics, time, skills, cost and resources.

Stage 7: Test design fit

Can a justified research design turn the available evidence into an answer to the question?

Stage 8: Narrow or restructure

Remove unnecessary populations, variables, outcomes or contexts. Split only when subquestions genuinely support one central inquiry.

Stage 9: Seek targeted feedback

Ask a supervisor or appropriate reviewer about the weakest decision in the chain rather than asking for general approval.

Stage 10: Freeze only the version needed for the next formal stage

Use the best current version for the proposal, ethics submission or data-collection plan, while recognizing that justified amendments may still be necessary later.

Final research-question audit

Before treating the question as proposal-ready, you should be able to answer yes to the following:

• I can explain the research problem separately from the question wording.
• The question is understandable without relying on vague concepts.
• The scope fits the degree level, available time and resources.
• The question is connected to current scholarship.
• I have not invented a research gap or originality claim.
• I know what evidence would be required to answer the question.
• I have a realistic route to that evidence.
• Ethical and governance requirements have been considered.
• The proposed design is justified by the question, not selected from keywords.
• The analysis could produce an answer that matches the question.
• The question does not assume the conclusion it is meant to investigate.
• Any subquestions directly support the central question.
• I am willing to revise the question if literature, access or feasibility changes.
• The wording and structure comply with my programme and disciplinary conventions.

A strong research question creates alignment. It connects the research problem, literature, evidence, methodology and practical constraints into one coherent project. The goal is not to produce a sentence that sounds sophisticated. The goal is to define an inquiry that can be defended, investigated and answered within the real conditions of the research project.