How to Write a Research Proposal

What a proposal has to prove

Strip away the section headings and every reader — admissions tutor, panel member, supervisor — is checking four things:

• What exactly are you asking? (A question, not a topic.)
• Why does it matter, and why has nobody answered it?
• How will you answer it, concretely enough to be criticised?
• Can you actually do it — in the time, with the access and skills you have?

Most weak proposals answer the first two well and the last two barely at all. Reviewers notice: a beautiful rationale attached to a vague method reads as enthusiasm, not a plan.

The eight sections

Names vary by discipline and funder, but the content is stable. Rough proportions assume a 2,000-word proposal.

1. Title and summary (~100 words). A working title that names the phenomenon, the population and the approach, plus a short abstract that states the question, method and expected contribution.
2. Background and problem statement (~250 words). Set the problem in the world, not in the literature. What goes wrong, for whom, at what scale?
3. Literature review and gap (~400 words). Short, synthesised and argumentative — see below.
4. Research questions or aims (~150 words). One main question and two or three sub-questions, or a set of numbered objectives. Number them; you will refer back to the numbers.
5. Methodology (~500 words). Design, participants or corpus, sampling, instruments, procedure, analysis. Say what you will do with the data once you have it, not just how you will get it.
6. Timeline and feasibility (~200 words). A table by month or quarter, with the milestones a reader can check you against.
7. Ethics and risk (~150 words). Which approval body, which route, how consent, anonymity and data storage are handled — and what you do if recruitment fails.
8. References. Full, consistent and real.

Test the question before you build on it

The most useful screen is FINER, introduced by Stephen Hulley and colleagues in Designing Clinical Research and now taught well beyond medicine. A good question is Feasible (subjects, expertise, time and money are within reach), Interesting to you and to the field, Novel — it confirms, refutes or extends what is known, Ethical, meaning a review board could approve it, and Relevant to knowledge, policy or future research.

Run your question through all five before drafting. A question that fails "feasible" costs you the whole proposal if you find out in month three.

The gap statement is the hinge

The literature review exists to earn one sentence. Compare:

Weak: "Many studies have examined remote work and productivity. However, more research is needed in this area."

Strong: "Studies of remote work and productivity have relied almost entirely on self-reported output from knowledge workers in large firms. Whether the same relationship holds where output is objectively logged — in service roles with measurable ticket volumes — has not been tested."

The strong version names the limitation, names the population it excludes, and makes the next study obvious. "More research is needed" names nothing, and every reviewer has read it a thousand times. If you are unsure whether your gap is real, that is a search problem before it is a writing problem — work through a systematic search of the literature first.

How long, and what the reviewers score

Length is set by the call, not by convention. PhD application proposals typically fall between 1,000 and 3,000 words depending on the department. Funder limits are hard numbers and strictly enforced:

• NIH (R01) — 1 page of Specific Aims, plus a 12-page Research Strategy organised under three required headings: Significance, Innovation and Approach.
• NSF — a Project Description of no more than 15 pages, with all charts, graphs and images counting toward the limit.
• UKRI — a core set of application questions rather than a single document: vision, approach, applicant and team capability, resources, and ethical and responsible research considerations.

The scoring criteria are usually published, and they are not always what you would guess. Every NSF proposal is assessed against two National Science Board criteria applied together — Intellectual Merit and Broader Impacts — and a proposal that fails to address both separately in the Project Summary and the Project Description is returned without review. Find your call's equivalent list and use its language as your subheadings.

Five mistakes that sink proposals

1. A topic instead of a question. "Social media and adolescent mental health" is a field. "Does reducing notification frequency change reported anxiety in 14–16 year olds?" is a project.
2. A literature review that summarises instead of arguing. Paragraphs that begin "Smith (2021) found…" one after another are a reading list. Group by finding, not by author.
3. Methods that cannot be criticised. "Interviews will be conducted and analysed thematically" tells a reviewer nothing. How many, recruited how, coded by whom, to what point of saturation?
4. A timeline with no failure branch. Say what happens if the archive is closed, the ethics decision is late, or recruitment stalls at half your target.
5. Citations you have not opened. Padding a proposal with references you have only seen in an abstract — or that an AI assistant supplied — is the fastest way to lose a reviewer who knows the field. Models invent plausible-looking references routinely, and a fabricated citation in a funding application is misconduct, not a typo.

Frequently asked questions

Keep reading

• How to Write the Methodology Section of a Research Paper
• How to Write a Literature Review Without Fabricating Sources
• Systematic vs Scoping vs Literature Review
• Free Citation Generator (APA, MLA, Chicago, Harvard, IEEE)