A research question is the single sentence that a study exists to answer. Everything downstream — the methodology, the sampling plan, the analysis, and ultimately whether reviewers and funders find the work compelling — depends on getting this sentence right before anything else is drafted. This guide walks through narrowing a broad topic into an answerable question, evaluating it against the FINER criteria, and adapting the process for clinical and health-related research using PICO/PICOT.
Start with a topic, not a question
Most researchers begin with a broad area of interest, not a question — “adolescent social media use,” “soil carbon sequestration,” “medication adherence in older adults.” A topic is not yet researchable: it doesn’t specify a population, a variable, a relationship, or a boundary on scope. The work of question formulation is narrowing that topic through a few concrete steps:
- Read for gaps, not just background. A literature review at this stage isn’t about summarizing what’s known — it’s about identifying what remains unanswered, contested, or untested in a new population or context. Look specifically for language in existing papers’ discussion or limitations sections flagging what wasn’t addressed.
- Narrow along one axis at a time. Pick a specific population, a specific variable or intervention, a specific outcome, or a specific context, rather than trying to narrow all four simultaneously. “How does social media use affect adolescents” is still a topic; “Is daily short-form video use associated with self-reported sleep latency in 13-17 year olds” is a question.
- State it as a question, out loud. A topic can be described in a phrase; a research question cannot — it requires a question mark and, implicitly, an answerable form (a comparison, a relationship, a prevalence, a mechanism).
- Distinguish the research question from the hypothesis. The question is what you’re asking; the hypothesis is your predicted answer, stated as a testable claim. Not every study requires a hypothesis (much qualitative and exploratory work is question-only), but every study requires a question.
The FINER criteria
Once a candidate question exists, it needs to be evaluated, not just admired. The most widely taught framework for this is FINER, introduced by Stephen Hulley and colleagues in Designing Clinical Research and now taught well beyond clinical research as a general-purpose checklist for question quality. FINER stands for:
- Feasible — Is there realistic access to an adequate number of subjects or data points? Is the required technical expertise available to the research team? Can it be completed within the time and budget actually available? A scientifically ideal question that no one can execute is not a usable research question.
- Interesting — Would answering it actually intrigue the investigator, peer researchers, and the relevant field? Motivation matters across a project that may run for years; a question the researcher isn’t genuinely curious about is a risk to completion, not just a stylistic concern.
- Novel — Does it confirm, refute, extend, or provide new information beyond what’s already established? Novelty doesn’t require inventing an entirely new phenomenon — replicating a finding in an understudied population, or testing whether an established relationship holds under new conditions, both satisfy novelty in the FINER sense.
- Ethical — Is it amenable to a study design that an institutional review board (or equivalent research-ethics body) would actually approve? A question that can only be answered by an unethical design needs to be reformulated before any proposal work begins, not after.
- Relevant — Does the answer matter to scientific knowledge, to clinical or policy decisions, or to the direction of future research? Relevance is what separates a merely answerable question from one worth the resources a study requires.
FINER is a diagnostic tool, not a formula — a question rarely scores perfectly on all five criteria on the first draft. Its practical use is to surface which dimension is weakest (usually feasibility or novelty) so it can be addressed by narrowing scope, adjusting the population, or doing a more targeted literature search before time is invested in a full proposal.
PICO and PICOT for clinical and health-related questions
In clinical, health services, and evidence-based-practice research, question formulation typically uses a more structured framework than FINER alone: PICO (Population, Intervention, Comparison, Outcome), first proposed by Richardson and colleagues in the mid-1990s as a way to build clearly answerable clinical questions, and its extension PICOT, which adds a Time element.
- P — Population/Patient: who is being studied? Defined precisely enough to be identifiable (age range, condition, setting), not just “patients.”
- I — Intervention: what is being done, given, or exposed — a treatment, a diagnostic test, an exposure, a policy.
- C — Comparison: what is the intervention being measured against — a placebo, standard of care, an alternative intervention, or no intervention at all. Not every question requires an explicit comparison, but naming one (or explicitly noting there isn’t one) sharpens the question.
- O — Outcome: what is actually being measured — a specific, ideally quantifiable endpoint, not a vague direction like “improvement.”
- T — Time (PICOT only): over what time frame is the outcome assessed? Adding a time frame is often what turns a PICO question into something a specific study design can actually be built around.
A related, broader framework, PICOTS, adds a Setting element and is used particularly in comparative-effectiveness and health-technology-assessment research; see CASRAI’s PICOTS Criteria dictionary entry for how that extension is defined and used.
Worked example: a broad topic of “exercise and diabetes” becomes, through PICOT: In adults aged 45-65 with type 2 diabetes (P), does a supervised 12-week resistance-training program (I), compared with usual-care physical activity advice (C), reduce HbA1c (O) over 6 months (T)? Every element in that sentence maps to a decision the eventual study design has to make — who is enrolled, what the intervention actually consists of, what it’s compared against, what’s measured, and when.
Broad research questions vs. narrow research questions
Neither a research question that’s too broad nor one that’s too narrow serves a study well:
- Too broad produces a question that can’t be operationalized into a single study design — it usually signals that the population, variable, or outcome hasn’t actually been chosen yet, only the general subject area.
- Too narrow produces a question so specific to one context, timepoint, or sample that the answer has little relevance beyond that single instance — this is the failure mode FINER’s “relevant” criterion is designed to catch.
The right level of scope is usually the narrowest question that still answers something the field, a funder, or a decision-maker actually needs to know — which is why narrowing and the FINER/PICOT checks are iterative, not a single pass: a question typically gets restated two or three times as feasibility or relevance problems surface.
From research question to proposal
A well-formed research question is the anchor for everything that follows it in a proposal or manuscript: it drives the specific aims, the choice of methodology, and the framing of the introduction. CASRAI’s guide to writing an NIH Specific Aims page and to Research Project Proposal Examples both assume a research question has already been settled before that drafting begins — this page is meant to be used before those, not alongside them.
Frequently asked questions
What makes a research question “good”?
A good research question is specific enough to be answered by a single, feasible study design, and meets the FINER criteria: it’s feasible with the resources available, genuinely interesting to the investigator and field, novel in some real sense, ethically achievable, and relevant to scientific knowledge or practice.
What is the difference between a research question and a hypothesis?
A research question is what the study sets out to answer; a hypothesis is a specific, testable prediction of that answer. Exploratory and qualitative studies commonly proceed with a research question and no formal hypothesis; hypothesis-driven quantitative studies require both, with the hypothesis derived from the question.
How many research questions should a study have?
There’s no fixed number, but most well-scoped studies center on one primary research question, sometimes supported by two or three secondary questions. A proposal built around many co-equal questions usually signals the topic hasn’t been narrowed enough yet.
Is PICO only used in clinical research?
PICO originated in evidence-based clinical practice, but the same structured logic — defining the population, the variable or intervention of interest, what it’s being compared against, and the outcome measured — is adaptable to other applied and social-science fields, even where the PICO acronym itself isn’t used by name.
Do I need a hypothesis before I have a research question?
No — the sequence runs the other direction. The research question is formulated first, from a narrowed topic; a hypothesis, where the study design calls for one, is then derived as a specific predicted answer to that question.
