What makes a good high school research question

A good research question is one you could be wrong about. That is the test almost every weak question fails: “how does social media affect teenagers” cannot be wrong, because it is not a claim about anything specific enough to check. Narrowing an interest into a question means adding constraints until being wrong becomes possible — a specific population, a specific measure, a specific comparison, a specific window of time.

The rest of this guide is how to do that deliberately.

The five tests: FINER

FINER is the standard checklist and it is standard because it works. A question is worth starting if it is Feasible, Interesting, Novel, Ethical and Relevant. Three of those decide whether a high school project finishes.

Feasible is the one that kills most projects. Can this be answered with the time, data, access and skill actually available? Not in principle — actually, by this student, this term. A question requiring a lab, a clinical population, proprietary data or a year of fieldwork is not feasible, however good it is.

Interesting means interesting to the student first. They will read forty papers about this and revise a draft four times. Nothing else sustains that.

Novel is where families over-reach. Novel does not mean nobody has studied the area. It means your specific combination has not been done: a new population, a new context, a newly available dataset, an old method applied somewhere it has not been applied. Field-first breakthroughs are not the bar and were never the bar.

Ethical is mostly about human subjects. If the project involves people — surveys, interviews, anything with minors — consent, privacy and data handling have to be thought through before collection, not after.

Relevant means someone other than the student would want the answer. Say who, in one sentence.

Narrowing, worked

The mechanics are the same every time. Start with the interest, then add: who exactly, what measure exactly, compared to what, over what period, using what data.

Interest: social media and mental health.

Too broad: Does social media harm teenagers?

Still too broad: Does Instagram use correlate with anxiety in high school students?

Answerable: Among students at three schools in one district, is self-reported time on image-based platforms associated with body-image scores on a validated scale, controlling for age and gender?

The third version can be wrong. It names a population, a measure that already exists and has been validated, and a control. It is also feasible: a survey, a scale from the literature, a spreadsheet.

Interest: economics and local business.

Too broad: How did the pandemic affect small businesses?

Answerable: Using public business-licence filings, how did the rate of new food-service registrations in one county change in the three years before and after 2020, compared with retail registrations over the same period?

That question is answerable entirely from public records — the single most underrated feature a high school research question can have.

Interest: literature.

Too broad: How is gender represented in Victorian novels?

Answerable: In three novels published between 1847 and 1853, how do the narrators describe women’s paid work, and how does that description change when the narrator is female?

Humanities questions are narrowed the same way: a defined corpus, a defined feature, a defined comparison.

The failure modes, named

The question that cannot be wrong. If every possible result confirms the thesis, it is a topic, not a question.

The question that needs someone else’s permission. Anything depending on a hospital, a company’s internal data, or an institution’s approval will stall. If the project cannot start without a gatekeeper, assume it will not start.

The question already answered well. Twenty minutes of literature searching before committing saves eight weeks. If the answer exists, either find the version of it that has not been done, or move on.

The question that is really five questions. A question with three “and”s in it is a paper the student cannot finish. Pick one.

The question whose data does not exist. Enthusiasm for a topic is not the same as access to evidence about it. Check the data source before committing to the question, not after.

Choose the method and the question together

The most reliable trick for a first project is to invert the usual order: look at what data you can actually get, and let that shape the question. Public datasets, government statistics, digitised archives, replication of a published analysis, systematic observation, a survey you administer yourself — each of these makes a whole family of questions feasible and rules out others. Doing research without a lab covers the main routes.

A question chosen in the abstract and then handed a method usually has to be renegotiated in week four. A question chosen alongside its method usually survives.

Test it in one paragraph

Before committing, write this down and read it aloud:

I want to find out whether [specific relationship] holds among [specific population or corpus], measured by [specific measure], using [specific data source], compared with [specific alternative]. If I am wrong, I will see [specific result].

If any bracket cannot be filled, that is the part still to narrow. If the last bracket cannot be filled, the question cannot be wrong, and the work has not started yet.

When to change the question

Early and without regret. A question that turns out to be unanswerable in week two is a normal, healthy outcome — it is what the first weeks are for. The same discovery in week eight is a crisis. Most of the value of an experienced mentor is compressed into this one judgement: knowing in week two what a student would find out in week eight, and saying so.

If the student has a broad interest and no idea how to narrow it, that is the ordinary starting position, not a problem — most scholars arrive that way. Tell us what they keep coming back to and we will work through the narrowing with them.

Published 19 August 2026. Written by the Meridian Scholars team.