How to Write a Research Paper Introduction, Paragraph by Paragraph

A tested skeleton gives you the article's bones. This post is about putting flesh on them — one paragraph at a time, using a single structure for every paragraph in the paper: a topic sentence that states the claim, two to five sentences that support it, and a bridge to what comes next.

I'm a clinical psychologist and statistician. I finished my PhD on schedule and have published around twenty papers, and I still consider myself a fairly average researcher — none of what follows is a talent trick. It's a structure, the same one I used to stop drowning in my own reading and start making arguments a reviewer could follow. Below I walk through how to write a research paper introduction section by section — importance, background, gap and research question, method, results — each with its own job and its own characteristic failure, and then the feedback loop that checks it all before you polish a single sentence.

How to write a paragraph

Read this:

Cognitive-behavioral therapy (CBT) is an effective treatment, and considered the most empirically supported, for child and adolescent anxiety. Notwithstanding the effectiveness of CBT, a substantial proportion of youth remain impaired following treatment, with meta-analyses showing 40% of youth at post-treatment and 50% at 12-month follow-up still meeting diagnostic criteria. Of particular concern, many youths do not maintain treatment effects. An estimated 48% of anxiety disordered youth relapse over long-term follow-up, with females and those meeting diagnostic criteria for social anxiety disorder posing the greatest risk... Several approaches aimed at improving outcomes have been investigated, with the majority focusing on increasing exposure practice and modifying the type and/or amount of family involvement. Several studies have also investigated the relative effectiveness of group vs. individual therapy... Finally, extensive research has also been conducted on school-based interventions...

This paragraph covers CBT effectiveness, relapse rates, gender moderators, long-term outcomes, exposure, family involvement, group versus individual therapy, and school-based prevention — and it is not finished. Eight distinct topics. A reader following it cannot tell what claim they are being asked to accept, because no claim is being made. It is a tour of the literature, not an argument.

Now read this:

Anxiety disorders are common in the developmental stage of adolescence, with a prevalence rate of 4–8%. Anxiety disorders during adolescence inhibit the ability to seek autonomy and enter adulthood because they negatively affect social interaction, the development of independent living skills, and educational outcomes. Furthermore, these impairments can continue into adulthood if left untreated. Given the prevalence and negative impact of anxiety disorders in adolescents, it is an important challenge to design interventions that provide short- and long-term effectiveness in routine-care clinical settings.

One claim: anxiety in adolescence is common and consequential. Three sentences of evidence. One bridge to the next argument. The reader knows exactly what they have been told and where they are going next.

The academic paragraph structure that produces the second and prevents the first is simple: topic sentence, two to five supporting sentences, optional bridge. The topic sentence names the one thing the paragraph is about — often taken straight from the skeleton. The supporting sentences prove, illustrate, or qualify it, and citations belong here, in service of the claim. The bridge connects to the next paragraph or returns to the main argument.

The discipline is that every sentence must serve the topic sentence. When you find yourself writing something interesting that does not, it belongs in a different paragraph — or nowhere. This is not a small constraint: after months of reading you will have accumulated more interesting material than any single paper can hold, and the temptation to include it is exactly what produces the first paragraph above.

I made this mistake at the start of my PhD. I sat down after six months of reading and presented what I had learned, roughly in the order I had learned it — coherent as a reading log, incomprehensible as an argument. Presenting everything you know is not the same as making a claim. Writing is not the output of reading; it is a separate act. You choose the claim first, then select the evidence that supports it. Everything else, however interesting, stays out.

Importance

Before a reader will care about your methods or results, they need to be convinced the topic is worth their time. That is the only job of the Importance paragraph — not to demonstrate how much you know, not to survey the literature, not to log every caveat. It answers one question: why should anyone care about this?

The claim takes one of four forms. Something is bad: harmful, costly, or impairing, and the numbers support it. Something is good: a treatment works, an approach shows promise, a resource is underused. Something is new: a problem has emerged, a gap has opened, the field has moved. Something is old and forgotten: a long-standing problem has not received the attention it deserves. That is the full menu, and your Importance paragraph picks one — occasionally two.

The ingredients are simple. Prevalence tells the reader how common the thing is; consequences tell them what it does. Together they establish that the topic exists and matters, and a bridge connects that to what the field is doing about it. In practice: anxiety disorders are common in adolescence, affecting 4–8% of young people; they impair the development of autonomy and adult functioning; without treatment, these impairments often persist. Three sentences — prevalence, consequences, chronicity — plus a fourth to bridge to the intervention literature. Nothing else.

The common mistake is treating this paragraph as a chance to show what you have read. My co-authors hit exactly this in an early draft of mine: the opening covered so much ground they could not tell what the article was about. It had surveyed everything and claimed nothing. When I rewrote it with a single claim and three pieces of evidence, the rest of the introduction suddenly made sense — modest enough to defend, clear enough to test. State why the topic matters: one claim, two or three pieces of evidence, a bridge.

Background

A Background section is not a literature review. It is preparation for your results. Everything in it should be there for one reason: a variable in your results depends on it.

So write it from the end. Before the first sentence of the introduction, list every variable you plan to assess in the results. Each one on that list needs coverage in the background; each one not on it probably does not belong there. Then the writing is mechanical: one paragraph per variable, same structure as any other — a topic sentence stating what is known, supporting sentences filling in prior effect sizes, key studies, established findings. The selection principle is simply: does this help the reader understand what you found? If yes, include it; if no, leave it out, however interesting.

Tone matters as much as selection. The background is matter-of-fact — CBT has shown consistent effects for youth anxiety across multiple meta-analyses — not strikingly, CBT has emerged as the gold standard. The striking framing is the Importance paragraph's job; here you just say what is known, plainly.

The failure mode is material that does not connect to your results, because reviewers read the background as a set of promises: every variable you introduce, they expect addressed. A background covering group therapy, family involvement, and school-based prevention sets up the expectation that the results address all three; if they address one, the reviewer's question is reasonable — if you knew these mattered, why did you not measure them? "I found them interesting while reading" is not a methodological defence. So: start with your results, list the variables, give each one a paragraph, state what is known plainly, then stop.

Gap and research question

A gap is not a revolution. It does not need to be the most important missing piece in your field's history. It needs to be one specific thing the background does not cover — small enough that your study can actually fill it.

The gap paragraph has a fixed structure. Open by naming what the background established: we know X, we know Y. Then name what is missing: despite this, we do not know Z. Then connect the gap to the importance: given [the reason the topic matters], there is a need to know Z. That is the gap paragraph, and it ends there. New paragraph. The research question: the current study examines Z, in this sample, using this design.

They belong in separate paragraphs because they do different work — the gap establishes that something is missing and matters, the research question says what you did about it. Mix them and you get a paragraph explaining what is known, what is not, and what you did all at once, and the reader loses the thread. The link to importance is what makes the gap more than a curiosity: "no one has studied this" is an observation, but "no one has studied this, and given that the problem is common and harmful, its absence has consequences" is a gap worth filling.

My own early draft buried the gap inside more background — more detail on which studies combined which strategies, another gap, a need statement, then the research question, all in two dense paragraphs — and my co-authors could not tell when the argument moved from what is known to what is missing to what we did. Giving the gap its own paragraph, ending cleanly, then opening a fresh one for the research question, gave it the weight it needed. The gap is one paragraph. The research question is the next. Keep them apart.

Method

The Methods paragraph in the introduction is the section researchers most often over-think. They write it as technical documentation — comprehensive, covering every design decision — and produce something three times longer than anything in their target journal; or they spiral over what "enough for reproducibility" means and produce something too thin.

The answer to both is the same: open your model article and read how they did it. This paragraph is not the full Methods section. It is a brief orientation — enough for the reader to grasp what kind of study this is, who the sample was, what was measured, and how the analysis worked. The right amount of detail is defined by your field, in the articles it publishes. Find the paragraph in your model article where they describe their design — usually one near the end of the introduction, after the research question — read it carefully, then rewrite it with your study's details. Same sentence structure, same specificity, same elements in the same order. Heavily inspired, not copied.

This is not laziness; it is precision. You are matching the conventions of a field that has, through years of peer review, settled on what it needs here. The failure mode is the attempt to improve on that — to be perfectly comprehensive, to write something that would stand alone as a methods description. That belongs in the full Methods section, where reviewers will look for it. The introduction paragraph is a signpost: short, structured, consistent with what the field considers normal — because normal, here, means correct.

Results

A Results section is not a data dump. It is a curated report of what you found — in the format your field expects, in the order that follows from your research question, stating plainly what the analysis showed.

The setup is already done: your analysis document has the numbers, the tables, the inline statistics. The task is to arrange them, which again means looking at your model article. How long are their results paragraphs? Do they lead with a table or with text? Do they report effect sizes alongside significance? Match that. The structure follows your research question — primary findings first, clearly labelled, with the statistic and the effect size; secondary findings next; sensitivity analyses and assumption checks last. Each finding gets one paragraph: a topic sentence stating what you found, the supporting statistics, and nothing else.

No interpretation — that is where most mistakes happen. The temptation is real, especially when findings are unexpected, but "this result may reflect..." and "one possible explanation is..." are Discussion sentences. State what happened, report the number, stop. The other failure mode is the data dump — every analysis, every check, no hierarchy, no clear primary finding. As a reviewer I find it hard to read, and the reaction is not charitable: a results section that presents everything without prioritisation looks like the writer does not know what matters, and that uncertainty invites scrutiny. The test is simple — after reading your results, could a reader state in one sentence what the study found? If yes, the section is doing its job.

Iterate with feedback

You have a full draft. The temptation is to send it to your supervisor and ask what they think. Resist it. General feedback on a full draft produces general comments: your supervisor responds to whatever catches their eye — a phrase, a table format, a word choice — and if the content is still incomplete, you have spent their attention on the surface of a draft nobody has checked underneath.

The alternative is the principle that governed the skeleton stage: targeted questions, to the right person, at the right moment. For the introduction and discussion, go to a field expert — not "what do you think?" but Is there a key reference I'm missing in the background on this variable? Is there an aspect of this mechanism I haven't accounted for? For the methods and results, go to your statistician or a methods-fluent colleague: Is there a step I've overlooked? A reporting convention I'm not following? Specific, answerable, aimed at the right expertise for that section.

What you are not asking, at this stage, is comment on prose or flow — that comes later. The only question now is content: is anything missing? A foundational citation, a variable you forgot to describe, a sensitivity analysis this journal's reviewers routinely expect. A targeted content check catches these; sentence-editing never will. A senior researcher who wants to help does exactly what is asked: ask for a full-draft review and you get general feedback; ask them to check the introduction for missing references on anxiety prevalence and they do that, well, because it is precisely what they know. The ask determines the outcome. So run the loop once per section — field expert for content, methods expert for method, both asked what is missing — then move forward.

The whole introduction, in one shape

Every paragraph in your article now has the same anatomy: one claim stated first, evidence in service of that claim, a bridge onward. Importance makes the reader care; background prepares exactly the variables the results deliver; the gap names what is missing and why it matters, and the research question, in its own paragraph, says what you did about it; method and results follow the field's conventions rather than your instincts. That is how to write a research paper introduction a reviewer can actually follow — not by writing better sentences, but by giving each paragraph one job and refusing to let it do anyone else's.

Which paragraph fights you hardest — importance, background, or the gap? Reply and tell me. I read every reply, and the patterns in what people get stuck on shape what I write next.