How to write a research question that you can actually answer

This guide is for students and early-career researchers who need a question for a thesis, a dissertation, a proposal or a review. It is the hub of our series on planning a study, so it links down to the hypothesis, methodology, framework and outline guides that come next.

What is a research question?

A research question is a single, answerable question that defines what your study will find out, about whom, and how you will know when you have an answer. Everything else in a thesis follows from it: the literature you read, the design, the sample, the analysis and the structure of your chapters.

A topic is not a question. "Social media and teenagers" is a topic. "Is daily time on short-video apps associated with sleep duration in 13 to 15 year olds in urban secondary schools?" is a question. The second version tells you who to sample, what to measure and which analysis to run.

What are the 3 types of research questions?

Most questions fall into one of three types, and the type decides your design.

Qualitative studies add a fourth kind in practice: exploratory questions that ask how people experience or make sense of something ("How do first-generation PhD students describe their relationship with their supervisor?"). These do not predict an outcome, which is why they rarely come with a hypothesis.

How to write a research question in four steps

Use this procedure. It takes most people a few sessions, not one.

1. Choose a topic you can live with. Pick an area you can read about for a year and where you can get data: access to a hospital, a school, a company dataset, a lab instrument or a public archive.
2. Read to find the gap. Read recent reviews and the most cited papers. Note what is unknown, contested or untested in your setting. A gap is usually one of: an unstudied group, an untested method, conflicting results, or an old finding in a new context.
3. Narrow to one sentence. Name the population, the main variable or phenomenon, and the outcome or context. Cut words such as "impact of", "role of" and "factors affecting" unless you can say which factors and how you will measure them.
4. Test and frame it. Run it through FINER (below), then write it in a framework (PICO, PCC or SPIDER) to check that every part is specific. Show it to your supervisor before you write the proposal.

Worked example: from topic to question (illustrative)

• Topic: mobile money and small businesses.
• Gap found: most studies cover Kenya and urban traders; few cover women-owned rural businesses in Tanzania.
• Draft question: "What is the impact of mobile money on women entrepreneurs?"
• Problems: "impact" is vague, no setting, no outcome.
• Final question: "Among women-owned rural retail businesses in two regions of Tanzania, is regular use of mobile money associated with higher monthly revenue after 12 months?"

What are the FINER criteria?

The FINER criteria are a five-point test of a research question: feasible, interesting, novel, ethical and relevant. They come from Hulley and colleagues' textbook Designing Clinical Research [1], and are now used well beyond medicine.

Feasibility kills more student projects than any other criterion. Before you commit, write down three numbers: how many participants or records you need, how many you can realistically reach, and how many months you have for data collection. If the second number is smaller than the first, narrow the population or change the design. Our statistical power and sample size guide shows how to get the first number.

How do you write a PICO question?

PICO is a framework for questions about the effect of an intervention or exposure: population, intervention, comparison, outcome. It was set out by Richardson and colleagues in 1995 as the anatomy of a "well-built clinical question" [2]. Adding T for time gives PICOT [3].

PICO worked example 1: health (illustrative)

Question: "In adults aged 65 and over living at home who fell in the past year, does a 12-week home-based balance and strength program, compared with usual care, reduce the number of falls over 12 months?"

PICO worked example 2: education (illustrative)

Question: "Do first-year statistics students taught with a flipped classroom score higher on the final exam than students taught with traditional lectures?"

PICO also works outside health whenever you compare an intervention. In engineering: "In reinforced concrete beams (P), does adding 1% steel fiber (I), compared with no fiber (C), increase flexural strength at 28 days (O)?"

What is the PCC framework?

PCC (population, concept, context) is the framework JBI recommends for scoping review questions, where the aim is to map evidence rather than test an effect [4]. It drops the comparison and outcome because a scoping review asks what has been studied, not what works.

PCC worked example: social science (illustrative)

Question: "What school-based mental health support programs for refugee children aged 5 to 12 have been described in high-income host countries?"

If you are choosing between review types, our scoping review vs systematic review guide has a decision table.

When should you use SPIDER instead?

SPIDER (sample, phenomenon of interest, design, evaluation, research type) is a framework for qualitative and mixed-methods questions, proposed by Cooke, Smith and Booth in 2012 as an alternative to PICO for qualitative evidence searches [5].

Example (illustrative), business: Sample, frontline retail managers; Phenomenon of interest, introducing self-checkout; Design, interviews; Evaluation, experiences and perceived effects on staff roles; Research type, qualitative. Question: "How do frontline retail managers experience the introduction of self-checkout, and how do they describe its effects on staff roles?"

Research question examples: five weak questions fixed

These are examples, written for this guide. Each shows the fault and the fix.

The pattern is the same each time: name the people, name the measure or phenomenon, name the setting, and remove words that judge before you have data.

How long should a research question be?

One sentence, usually 15 to 35 words. If yours runs longer, check for two questions joined by "and". Keep one as the main question and turn the others into sub-questions. A thesis commonly has one main question and two to four sub-questions, each answered in a results section or chapter.

How to write research questions with AI for Research

AI for Research has a Research question module in the Thesis Suite that turns a topic into candidate questions and checks them. Use it as a starting point, then apply the steps above.

1. Find the gap first. Search in Find Papers with filters for year, study type and peer-reviewed only, or use Landscape to see works per year and top topics for your area. The research gap finder suggests gaps, each tied to papers (3 free runs a day without an account; 3 credits when signed in).
2. Refine the question (3 credits). In the Thesis Suite, open Research question, enter your topic or draft question, add your setting and resources as context, and choose FINER, PICO, PICOT or SPIDER. You get one refined question, a table showing how it meets the framework, three alternative versions (narrower, broader, a different angle), what the literature says with citations, and three or four sub-questions.
3. Check novelty (0 credits). Paste your working title or a draft abstract into Similar papers to see the closest published work.
4. Move on (3 credits each). The Hypotheses, Methodology builder and Thesis outline modules take the question as their input.
5. Share it with your supervisor. Put the work in a Project and invite your supervisor as "can comment".

The free plan includes 45 credits a month, enough for several modules. Search, filters, Landscape and Similar papers cost nothing. See pricing. Limits: the tool cannot judge feasibility for your setting (it does not know your access, budget or deadline), and your university may expect a specific format, so check with your supervisor before you submit.

FAQ

What makes a good research question?

A good research question is focused, answerable with data you can get, and worth answering. It names who or what you will study, what you will look at, and what result would count as an answer. The FINER test sums this up: it should be feasible, interesting, novel, ethical and relevant. If you cannot picture the table or theme list that would answer it, it is not ready.

What are the 3 types of research questions?

Most research questions are descriptive, comparative or relational, or causal. A descriptive question asks what exists or how much (what share of nurses report burnout). A comparative or relational question asks how groups differ or how variables move together. A causal question asks whether one thing changes another, which usually needs an experiment or a strong quasi-experimental design. The type you choose decides your design.

How long should a research question be?

One sentence, usually 15 to 35 words. It must be long enough to name the population, the main variable or phenomenon, and the outcome or context, and short enough to read in one breath. If it needs two sentences or several 'and' clauses, you probably have two questions. Split them into a main question and sub-questions.

What is the difference between PICO and PCC?

PICO (population, intervention, comparison, outcome) fits questions about the effect of a treatment or exposure, so it suits trials and systematic reviews of effectiveness. PCC (population, concept, context) comes from JBI guidance for scoping reviews, where the aim is to map what is known rather than measure an effect. Use PICO when you compare; use PCC when you explore or map.

Can AI write my research question?

AI can suggest candidate questions from a topic and show which papers already address them, which saves time at the start. It cannot know your supervisor's expectations, your access to data or your deadline, so treat its output as a shortlist. Test each candidate against FINER yourself, check the papers it cites, and rewrite the one you keep in your own words.

Next steps

Your question leads to the next four planning steps:

• Turn it into testable predictions: how to write a hypothesis.
• Choose the theory behind it: theoretical vs conceptual framework.
• Plan how you will answer it: how to write a methodology chapter.
• Map the whole thesis around it: how to write a thesis outline.

If the question is for a proposal, continue with how to write a research proposal.

Sources

1. Hulley, S. B., Cummings, S. R., Browner, W. S., Grady, D. G., & Newman, T. B. (2013). Designing Clinical Research (4th ed.). Lippincott Williams & Wilkins. FINER criteria in chapter 2.
2. Richardson, W. S., Wilson, M. C., Nishikawa, J., & Hayward, R. S. (1995). The well-built clinical question: a key to evidence-based decisions. ACP Journal Club, 123(3), A12–A13. https://doi.org/10.7326/ACPJC-1995-123-3-A12 (PMID 7582737)
3. Riva, J. J., Malik, K. M. P., Burnie, S. J., Endicott, A. R., & Busse, J. W. (2012). What is your research question? An introduction to the PICOT format for clinicians. Journal of the Canadian Chiropractic Association, 56(3), 167–171. https://pmc.ncbi.nlm.nih.gov/articles/PMC3430448/
4. Pollock, D., Peters, M. D. J., Tricco, A. C., Munn, Z., et al. Scoping reviews. In E. Aromataris et al. (Eds.), JBI Manual for Evidence Synthesis. JBI, 2024. https://doi.org/10.46658/JBIMES-24-09 ; chapter online: https://jbi-global.atlassian.net/wiki/spaces/MANUAL/pages/355862497/10.+Scoping+reviews
5. Cooke, A., Smith, D., & Booth, A. (2012). Beyond PICO: the SPIDER tool for qualitative evidence synthesis. Qualitative Health Research, 22(10), 1435–1443. https://doi.org/10.1177/1049732312452938