A research question is the one question a project sets out to answer. A good one is narrow enough that evidence can settle it and open enough that the answer is not a fact you could look up. "Is social media bad for teenagers?" is a topic with a question mark. "Among 13- to 15-year-olds in two school districts, is time spent on short-form video after 9 p.m. associated with shorter weeknight sleep?" is a research question.
What makes a research question strong
Writing centers and methods textbooks list much the same qualities. Run every draft through these six.
| Quality | Ask yourself | It fails when |
|---|---|---|
| QualityFocused | Ask yourselfCan I name the population, the place and the time? | It fails whenIt says "teenagers" or "the world" |
| QualityResearchable | Ask yourselfIs there evidence I can get: data, sources, people? | It fails whenIt asks "should", or needs data nobody will share |
| QualityOpen | Ask yourselfDoes the answer take analysis, not a lookup? | It fails whenOne fact or one statistic settles it |
| QualityClear | Ask yourselfWould two readers define every term the same way? | It fails whenIt leans on "effective", "success" or "impact" |
| QualityFeasible | Ask yourselfCan I answer it with my time, skills and budget? | It fails whenIt needs ten years of data for a one-semester paper |
| QualityWorth answering | Ask yourselfWould anyone change a belief or a decision? | It fails whenThe answer matters only to you |
Health research uses a version of this list called FINER, from Designing Clinical Research by Stephen Hulley and colleagues: feasible, interesting, novel, ethical, relevant. The ethical test applies well beyond medicine. If the only way to answer your question is data an ethics board would not approve, such as students' private messages, change the question.
Weak vs strong research questions
Each pair starts from a typical first draft. The repair is usually one of four moves: add a population, add a place and time, swap a vague word for something observable, or split two questions apart.
Weak: How does climate change affect agriculture?
Strong: How have tart cherry growers in northwest Michigan changed the way they protect blossoms from spring frost over the past twenty years, and what decides which method they use?
Weak: What percentage of US adults have a library card?
Strong: What leads card-holding adults in a mid-sized city to stop visiting the library, and what brings some of them back?
Weak: Should schools ban phones?
Strong: In middle schools that require students to lock phones in pouches for the whole day, how did teachers' reports of classroom disruptions change over the first school year?
Weak: Is remote work effective?
Strong: On fully remote software teams, how does the number of time zones a team spans relate to how long code changes wait for review?
Weak: How do social media and video games affect teenagers' grades and mental health?
Strong: Among ninth graders in one district, is weekday gaming time associated with first-semester grades after accounting for eighth-grade grades?
Weak: How has the harmful rise of fast fashion damaged local tailors?
Strong: How have independent alteration shops in one city changed the services they offer over the past fifteen years, and how do their owners explain the change?
Weak: What did my 10 a.m. statistics section think of last Tuesday's quiz?
Strong: In introductory statistics sections at one university, do students who take weekly low-stakes quizzes score higher on the final exam than students who take monthly ones?
Weak: How do US hospitals use AI in diagnosis?
Strong: How do radiologists at three regional hospitals decide whether to accept or override an AI tool's flag on a chest X-ray?
Research question examples by type
The type of question decides the method. Descriptive questions need counting, causal ones need a comparison group, interpretive ones need close reading. Here is one topic, cycling to work in a mid-sized city, asked seven ways.
| Type | It asks | Example |
|---|---|---|
| TypeDescriptive | It asksWhat is happening, how much, how often | ExampleWhat share of commuters cycle to work weekly, and how does that change between summer and winter? |
| TypeComparative | It asksHow groups or places differ | ExampleHow do cycling rates differ between neighborhoods with protected lanes and those with painted ones? |
| TypeRelational | It asksWhether two things move together | ExampleIs distance to the nearest protected lane associated with how often residents cycle to work? |
| TypeCausal | It asksWhether changing X changes Y | ExampleDid bike counts on one street rise after a protected lane opened, compared with similar streets without one? |
| TypeExploratory | It asksHow people experience or explain something | ExampleHow do new bike commuters describe their first month, and what makes some of them stop? |
| TypeEvaluative | It asksWhether a program met its goal | ExampleDid the city's e-bike rebate reach the low-income households it was designed for? |
| TypeInterpretive | It asksHow something is represented or understood | ExampleHow do local news stories frame conflicts between drivers and cyclists? |
Quantitative, qualitative and mixed methods
Quantitative questions name variables and a relationship between them. Qualitative questions usually open with how or what and leave room for answers you did not expect. Many methods texts advise against "why" here, since it hints at cause and effect. A mixed-methods study asks one of each and links them.
Quantitative: Among adults aged 65 and over in one county, is living alone associated with fewer primary care visits a year, after adjusting for income and chronic conditions?
Qualitative: How do adults over 65 who live alone decide when a health problem is worth a visit to the doctor?
Mixed methods: Ask both, and use the interviews to explain the pattern in the visit data.
PICO and SPIDER for health questions
Clinical and nursing research often writes questions to a template. PICO, set out by W. Scott Richardson and colleagues in ACP Journal Club in 1995, stands for population (or patient), intervention, comparison and outcome. PICOT adds time. For qualitative evidence, Alison Cooke, Debbie Smith and Andrew Booth proposed SPIDER in 2012: sample, phenomenon of interest, design, evaluation and research type.
In adults aged 65 and over sent home after hip fracture surgery (P), does a nurse's phone call within 72 hours of discharge (I), compared with written instructions alone (C), reduce hospital readmissions (O) in the first 30 days (T)?
What do interview and focus-group studies (D) using qualitative methods (R) report about the experiences (E) of adult children (S) caring for a parent in the early stages of dementia (PI)?
Research question examples by field
Each set is scoped the way a class paper, thesis or article would scope it, with a note on what the questions share.
Education
- How do ninth-grade English teachers in one district change their take-home essay assignments after the school adopts an AI-use policy?
- Do students in districts on a four-day school week score differently on state reading tests from comparable students on a five-day week?
- What do first-generation college students say helped or hindered them in their first semester of online courses?
Psychology
- Among first-year undergraduates, is phone use after 10 p.m. associated with self-reported sleep quality and next-day mood?
- Does ten minutes of writing about an upcoming exam lower test anxiety scores compared with a neutral writing task?
Nursing and public health
- Is distance to the nearest pharmacy associated with how consistently adults over 60 in rural counties refill blood pressure medication?
- How do community health workers decide which households to visit first when their caseload exceeds their hours?
Sociology and political science
- How do food delivery couriers in Chicago decide which hours to work and which orders to accept?
- When a county loses its last local newspaper, does the number of candidates running for county office change in the following elections?
Business and economics
- How did small retailers inside Manhattan's congestion pricing zone, which began charging drivers on January 5, 2025, change their delivery schedules in the toll's first year?
- Do subscription businesses that offer customers a pause option keep them longer than businesses that offer only cancellation?
Environmental science
- How does tree canopy cover on a city block relate to its peak afternoon surface temperature in July in Phoenix, Arizona?
- How do coastal homeowners offered a government buyout after repeated flooding decide whether to accept it?
Computer science and technology
- Do novice programmers fix planted bugs faster with an AI code assistant than without one, and can they explain the fix equally well afterward?
- How do blind and low-vision people using screen readers work around inaccessible checkout pages on shopping sites?
History
- How did Philadelphia's daily newspapers describe the influenza epidemic in the fall of 1918, and did their tone change as deaths rose?
- What do the surviving letters of women homesteaders in Nebraska reveal about how their households managed debt?
Literature and media
- How do Robert Walton's letters, which open and close Mary Shelley's Frankenstein (1818), shape how far a reader trusts Victor's account?
- In Dickens's Bleak House (1852–53), what does Esther Summerson's first-person narrative notice that the present-tense narrator does not, and why does the novel need both?
- How do book recommendation videos on TikTok describe the novels they promote, and which features (plot, tropes, the viewer's emotional reaction) come up most often?
Walton's letters make Frankenstein a classic frame story, and questions about narrators and frames suit literature papers because the evidence is all on the page. For an argumentative essay, carry the question into an essay outline and let your thesis statement answer it.
Sub-questions, hypotheses and thesis statements
A thesis or dissertation usually has one main question and a few sub-questions, each mapped to a chapter or a strand of analysis. Together they should answer the main question. How many you need varies by field and program.
Main question: How did converting downtown office buildings to apartments in one mid-sized US city between 2020 and 2025 affect street-level shops on the surrounding blocks?
- Which buildings were converted, when, and how many homes did each add? (Descriptive, from permit records.)
- How did storefront vacancy on nearby blocks change compared with similar blocks without conversions? (Comparative, from vacancy data.)
- How do shop owners near converted buildings describe changes in their customers? (Exploratory, from interviews.)
In an IMRaD paper, the question closes the introduction and the methods section explains how you answer it (see this methodology section example). For a longer project, the guide to writing a dissertation shows how the question shapes each chapter.
The question comes first. A hypothesis predicts its answer, and a thesis statement argues for the answer once you have the evidence.
| What it is | Example | |
|---|---|---|
| Research question | What it isWhat you ask before you know | ExampleIs distance to a protected bike lane associated with how often residents cycle to work? |
| Hypothesis | What it isA testable prediction, mostly in quantitative work | ExampleResidents who live within 400 meters of a protected lane cycle to work more often than residents who live farther away. |
| Thesis statement | What it isThe answer you argue for | ExampleProtected lanes raise cycling mainly among people who live a short walk from them, which favors a dense network over a few long routes. |
Research questions for books and articles
A nonfiction author, a reporter on a long feature or a novelist building a historical setting gets more from a short list of real questions than from a note that says "research the 1880s". The same rules apply: specific, answerable, tied to what a chapter needs.
- Historical novel: What did a lighthouse keeper on the Maine coast do in a working day in the 1880s, hour by hour, and what do keepers' logs record?
- Memoir: What work did people in my grandmother's village find in the ten years after the mill closed?
- Feature story: Why did three bakeries on one street close within a year, and what do their leases and their former owners say about rent?
In Strokes, a book is a project folder with chapters as documents, so the questions can sit beside the draft as a to-do list. Add that document to a chat with the AI Assistant and ask it to test each question against the six checks above. Use it as a sparring partner, not a source: language models can cite papers that do not exist (an AI hallucination), so look up every reference yourself.
How to write your own research question
- Start from a topic you already know a little about
Read two or three recent review articles or a textbook chapter to see what is settled and what is still argued over.
- Write down every question the reading raises
Aim for ten rough questions. In recent papers, phrases such as "remains unclear" or "future research should" mark gaps others have noticed. Note sources as you go. An annotated bibliography saves you from finding them twice.
- Pick one and add boundaries
Add who, where, when and which outcome, in one sentence ending in a question mark. If it needs a semicolon, it is probably two questions.
- Match the wording to a method you can carry out
Decide what evidence would answer it: a dataset, a survey, interviews, an archive, a set of texts. If you cannot get it, change the question.
- Write down the answer you expect
Draft a hypothesis or working thesis. If only one answer seems possible, the question is too simple or leading. If none does, it is too vague. Then run the checklist below, and expect the question to shift once you have read more.
Once the question holds up, build the paper around it with a research paper outline, and when the work is done, state the question in your abstract in a single plain sentence.
Frequently asked questions
How many research questions should a paper have?
A class paper usually needs one. A thesis or dissertation commonly has one main question and a few sub-questions, each answered by a chapter or a strand of analysis. Expectations differ by field and institution, so check the assignment brief or ask your supervisor.
Can a research question be a yes-or-no question?
In quantitative research, yes, when the yes or no comes from a test: 'Does a nurse's follow-up call reduce readmissions?' is a standard causal question. What fails is a yes-or-no question that one fact settles, such as whether a law exists. In qualitative and humanities work, how and what questions usually serve better.
Where does the research question go in a paper?
In fields that use the IMRaD structure (introduction, methods, results, discussion), at the end of the introduction, after the background and the gap in existing work. In humanities essays it usually appears early in the introduction as the problem your thesis statement answers.
