How do I narrow down a research question that is too broad?

Broad questions feel safe because they cannot exclude anything, which is exactly the problem. A question that admits all evidence gives you no basis for deciding what to read, what to collect, or when you are finished. Most of the paralysis that gets described as procrastination in year one is a scoping problem wearing a motivational disguise.

The instinct when told to narrow is to add qualifiers everywhere. That produces the over-specified question: informal mentoring, among early-career teachers, in rural secondary schools, in one region, during a three-year window, measured by intention to leave. Each qualifier was reasonable and together they define a study nobody was asking for.

Cutting one dimension decisively works better because the question stays legible. "Does mentoring affect retention" becomes "does mentoring affect retention in the first three years", and the field can still see what is being asked. The other dimensions remain broad, which keeps the finding transferable, which is what makes it worth publishing.

Narrowing works better as a sequence than as a single edit. Ratan and colleagues describe starting from the broad subject, doing preliminary reading, writing down what is already known and what is still unknown, then rewriting the general question as progressively more specific ones. Their worked example moves from hormone levels in hypospadias to differences in reproductive hormone levels between children with isolated hypospadias, children with associated anomalies, and a normal population (Ratan et al., 2019, p. 18). The order that produces that result is topic, existing knowledge, missing knowledge, population, variables, outcome.

Test the narrowed version against feasibility before you commit to it. Ratan and colleagues define feasibility through available subjects, an appropriate method, time, funds, access to people or documents, and the ability to tie your concepts to observable indicators, and they note that a question can look feasible until fieldwork starts (Ratan et al., 2019, p. 15). Their remedies map onto the failure: specify fewer variables when the question is still too broad, reduce subjects or measurements when costs are excessive, collaborate or acquire the skill when the method exceeds what you can do, and consult a supervisor or peers when relevance is the uncertain part (Ratan et al., 2019, p. 16).

Choose the dimension by asking where the interesting variation lives. If everyone assumes the effect is uniform across settings and you suspect it is not, narrow on setting and say why. The narrowing is then part of the argument rather than a concession to feasibility.

Which parts of a question can I narrow on?

Six levers. Population, setting, time period, the outcome you measure, the mechanism you examine, and the theoretical lens you apply. Write your question six times, narrowing one lever each time, and compare. The best version is usually obvious once they sit side by side.

Mechanism is the lever most often overlooked and frequently the most productive. Moving from "does X affect Y" to "does X affect Y through Z" narrows the question without shrinking who cares about it, because a mechanism claim speaks to everyone studying the relationship.

Theoretical lens is the lever to use with most caution. Narrowing to one framework can turn a substantive question into an exercise in applying a theory, which is a thinner contribution and harder to defend.

How do I know when I have narrowed too far?

When you cannot say who besides your examiners would want the answer, or when the answer is obvious before you collect anything. Both mean the question has stopped being research. Widen the lever you cut last, since that is usually the one cut furthest.

The clearest sign of over-narrowing is that the question now returns a fact instead of a relationship. Ratan and colleagues contrast "How much time do young children in Delhi spend doing physical activity per day?", which produces a statistic, with "What is the relationship between physical activity levels and childhood obesity?", which produces something to analyze. They set the opposite boundary with the same pair of examples, where "What are the effects of protein-energy malnutrition in schoolchildren in New Delhi?" is unfocused and "How does protein-energy malnutrition among children affect academic performance in elementary schoolchildren in Delhi?" names a population and an outcome (Ratan et al., 2019, p. 16). Cut variables and cases that do not serve the central relationship, and stop before the question can be answered with a number or a yes.

A second symptom is a literature review with nothing in it. If almost no prior work addresses your question, that occasionally signals a genuine gap and much more often signals that you have narrowed past the point where anyone was interested.

The most useful check is a conversation. Describe the question to a researcher in your field and watch whether they ask a follow-up. Questions worth answering generate questions in return. Over-narrowed ones generate polite agreement.

What you narrow is the evidence and the analysis, not the significance. Pautasso qualifies the advice to focus by telling authors to balance a narrow scope against broader relevance, and to spell out the implications for neighboring disciplines rather than leaving them implicit (Pautasso, 2013, p. 3). Pautasso also treats the search itself as part of the boundary, recommending that you record your search terms so the search can be repeated, define exclusion criteria early, and read previous reviews alongside primary studies (Pautasso, 2013, p. 1). Write the one-sentence gap statement first, then use it to reject the studies, populations, and periods that do not help you answer it.