How to write a research question
Short answer: A research question is the specific, answerable question your study is designed to answer. A good one names the population or context, the key concepts or variables, and the relationship or phenomenon you will examine, and it can be answered with data you can realistically collect. To write one, choose a topic, read enough to find a gap, narrow the scope, draft the question with a clear question word, and test it against the FINER criteria (feasible, interesting, novel, ethical, relevant); in clinical and health research, structure it with PICO (population, intervention, comparison, outcome).
On this page
- What a research question is (and is not)
- Types of research questions
- Step 1: Start from a topic you can read about
- Step 2: Read to find a gap
- Step 3: Narrow the scope
- Step 4: Draft the question
- Step 5: Test it against FINER (and PICO where it fits)
- The FINER criteria
- PICO for clinical and health questions
- From weak to strong: examples
- How many research questions?
- Common mistakes
- Quick checklist
Almost every decision in a study follows from its research question: what you read, what data you collect, which methods you use and what counts as an answer. A vague question produces a vague paper. A precise one tells you, and your readers, exactly what the study is for.
What a research question is (and is not)
A research question is a single, focused question that your study is designed to answer. It is narrower than a topic ("remote work") and different from a hypothesis, which is a testable prediction about the answer ("remote-work intensity is positively associated with exhaustion"). Many quantitative studies state both: the question frames the study, and the hypotheses specify what the data are expected to show. See how to write a hypothesis for the next step.
Types of research questions
| Type | What it asks | Typical design | Example |
|---|---|---|---|
| Descriptive | What is happening, how much, how often? | Survey, descriptive statistics | What proportion of first-year students work more than 15 hours a week? |
| Comparative | How do groups differ? | Group comparison, experiment | Do online and in-person students differ in course completion rates? |
| Relational | How are variables associated? | Correlational, regression | Is weekly paid work associated with lower grades among first-year students? |
| Causal | Does X change Y? | Experiment, quasi-experiment | Does a weekly feedback session improve first-year writing scores? |
| Exploratory (qualitative) | How or why do people experience something? | Interviews, case study | How do first-year students who work part time manage study deadlines? |
The type you choose shapes the method. A causal question needs a design that can support causal claims; if you can only collect cross-sectional survey data, ask a relational question instead.
Step 1: Start from a topic you can read about
Pick an area that interests you and that has an existing literature. "Student employment" or "remote work and wellbeing" is a topic, not yet a question. Write down what you already know and what puzzles you.
Step 2: Read to find a gap
Skim recent reviews and a handful of recent studies. Look for findings that conflict, populations or settings that have not been studied, outcomes nobody measured, or methods with obvious weaknesses. The gap is what makes your question worth asking; our guide to literature reviews shows how to map it.
Step 3: Narrow the scope
Add boundaries until the question could be answered by one study: a population, a setting, a time frame, a specific outcome. "How does work affect students?" becomes "Is weekly paid work associated with lower first-semester grades among full-time undergraduates at public universities?"
Step 4: Draft the question
Use a question word that fits your aim: what or how much for description, is there a relationship or to what extent for association, does for causal questions, how or why for exploratory qualitative work. Name the key concepts in terms you can measure or observe. Avoid yes/no wording when what you really want is a magnitude ("to what extent" rather than "does … at all").
Step 5: Test it against FINER (and PICO where it fits)
Check the draft against the FINER criteria, introduced by Hulley and colleagues in Designing Clinical Research and now used well beyond medicine. Revise until it passes; this is usually where a question gets narrower.
The FINER criteria
| Criterion | Ask yourself |
|---|---|
| Feasible | Can I get enough participants or data, with the skills, time and money I have? Is the scope manageable? |
| Interesting | Does the answer matter to me, my supervisor and people in my field? |
| Novel | Does it provide new findings, or confirm, refute or extend earlier ones (a replication in a new population counts)? |
| Ethical | Can it be done without unacceptable risk, and will an ethics committee approve it? |
| Relevant | Could the answer affect knowledge, practice, policy or future research? |
PICO for clinical and health questions
In clinical and evidence-based health research, questions are often built with PICO, a structure described by Richardson and colleagues in 1995 ("The well-built clinical question"):
| Element | Meaning | Example |
|---|---|---|
| P | Population or problem | Adults with type 2 diabetes |
| I | Intervention or exposure | A smartphone self-monitoring app |
| C | Comparison | Usual care |
| O | Outcome | Change in HbA1c after six months |
Combined: "In adults with type 2 diabetes, does a smartphone self-monitoring app, compared with usual care, reduce HbA1c after six months?" Variants add T (time frame) or S (study design). For qualitative questions, some fields use similar frameworks that replace the intervention and comparison with the phenomenon of interest and context.
From weak to strong: examples
| Weak | Why | Stronger |
|---|---|---|
| What are the effects of social media? | Too broad; no population or outcome | Is daily time on short-video apps associated with sleep duration among 13- to 15-year-olds? |
| Is remote work good? | Value judgment; not measurable | To what extent is remote-work intensity associated with emotional exhaustion among software employees, and does boundary management explain this association? |
| Why do companies fail? | Too broad for one study | How do founders of failed food-delivery start-ups in two cities explain the decision to close? |
| Does tutoring help? | Missing population, comparison and outcome | Does weekly peer tutoring, compared with no tutoring, improve end-of-term mathematics scores among first-year engineering students? |
How many research questions?
Most journal articles and theses have one main question, sometimes with two to four sub-questions that break it into parts. Sub-questions should each be answerable and together should answer the main question. If you have more than four, the study is probably trying to do too much.
Common mistakes
- Asking a question that is really a topic or a title
- Building the answer into the question ("How harmful is…")
- Using concepts you cannot measure or observe
- Asking a causal question with a design that cannot support causal claims
- Choosing a question with no data you can realistically obtain
- Changing the question after seeing the results without saying so
Quick checklist
- Is it a single, clear question with a question mark?
- Does it name the population or context and the key concepts?
- Is it neither so broad that it needs a book nor so narrow that the answer is trivial?
- Is the type (descriptive, relational, causal, exploratory) matched to a feasible design?
- Does it pass the FINER criteria?
- Does it address a gap you can point to in the literature?
- Can you imagine the table, figure or theme that would answer it?
Once the question is settled, it becomes the backbone of your research proposal and of the introduction to your paper.