Research Paper Limitations, Conclusion, and the Final Polish

Your Discussion is written. You've summarised your findings in plain language and interpreted them one matched pair at a time, inside a closed loop. (If you haven't done that yet, start with the companion post on [how to write a discussion section] — this picks up exactly where it ends.)

What's left is the part that decides whether the paper survives peer review and whether anyone ever finds it: the limitations, the conclusion, and the packaging. This is where careful researchers lose their nerve — they hedge too much, conclude too weakly, and rush the details. Let's not.

I'm a clinical psychologist and statistician. I finished my PhD on schedule, I've published around twenty papers, and I'd describe myself as a fairly average researcher — which is the point. What follows isn't talent. It's a fixed order of small steps, and here it is:

1. Address the three limitations a reviewer would raise — before they raise them.
2. Conclude with conviction, not hedges.
3. Then, and only then, write the keywords, title, and abstract.
4. Make one final pass for all the small details you've been deliberately deferring.

By the end, the paper isn't almost done. It's done.

Step 1: Limitations — answer the three objections before the reviewer does

A limitations section is not a confession. It's a preemptive response to the three objections a knowledgeable reviewer is most likely to raise — written before they can raise them, with a reasoned case for why your findings are still interpretable.

Three limitations. Not two, not five, not a comprehensive inventory of everything that could have been done differently. Three — the most substantive issues a reviewer will notice, handled before they ask.

Each one follows a fixed structure. You name it plainly. You explain why it was hard to avoid — inherent to the design, the population, the setting, or the resources. You assess how it might affect the findings. You describe what you did to mitigate it. And you close with a brief, reasoned case for why the findings still stand.

Here it is applied:

The study's sample was drawn from a single outpatient clinic, which limits generalisability to other settings. This was a consequence of the study's effectiveness design, which required a naturalistic treatment context. However, participants were recruited consecutively without exclusion criteria based on severity or comorbidity, and the sample's diagnostic composition was broadly representative of clinical populations in comparable settings.

See what that does: it names the limitation, explains why it was unavoidable given the design goals, and shows the threat to validity is smaller than it first appears. The concern is acknowledged and addressed. The finding is left standing.

Two things should never appear in a limitations section.

A limitation you could have fixed. Noting that a larger sample would have increased power — when nothing actually prevented you from recruiting more participants — isn't honest self-assessment. It's an admission that something avoidable wasn't avoided, and reviewers read it exactly that way. If you genuinely couldn't recruit more for clinical or logistical reasons, say so and explain why.

A limitation with no mitigation and no relevance. A weakness listed because it exists, not because it matters. A study without a control group has a limitation — but if your design was explicitly non-comparative (a single-arm effectiveness study), the absence of a control group is a design choice, not a weakness. Listing it confuses the reader about what your study was even trying to do.

The common failure runs in both directions. Some writers list so many limitations the paper sounds worthless — every design decision, every measurement constraint, until the reader wonders why it was published at all. Others list nothing, which signals either naivety or evasion, and guarantees reviewers will fill the gap themselves with objections your paper isn't prepared to answer.

And when a reviewer does raise an objection anyway — as they will — that's a separate skill: responding without either caving or getting defensive. I've written about that elsewhere.

Three limitations. Genuinely hard to avoid. Assessed for impact. Mitigated as far as possible. The purpose is context — not self-criticism, not a performance of humility. You're helping the reader interpret what you found.

📄 Free: The small stuff — formatting, terminology, journal conventions — is exactly what derails a final draft. Grab the 20-Point Final Polish Checklist: a one-page pre-submission sweep you run once, right before you submit, so none of it lives in your head while you write. Download it here.

Step 2: Conclusion — take a stand, then stop

The conclusion is the last thing a reader holds in their mind when they close your paper. It deserves more than a trailing sentence about future research.

What separates the conclusion from the Discussion summary is stance. The summary reported what happened. The conclusion takes a position on what it means — and points to what should happen next. It isn't neutral. It commits.

Here's what that looks like:

In conclusion, this trial provides support for the use of multi-family, multi-disorder group CBT for adolescent anxiety disorder that includes high exposure to feared situations and high levels of parental and school involvement. A particularly promising result was that only one of the participating adolescents who achieved remission at post-treatment relapsed during the follow-up period, and many participants who had not achieved remission at post-treatment achieved remission during the follow-up. Furthermore, it provides proof-of-concept that this approach is feasible within routine-care clinics and effective across a range of included diagnoses. Further research should evaluate the described approach in a randomised controlled design to further investigate its potential in a stepped care approach.

Four sentences: a claim ("provides support for"), a highlighted finding (the relapse rate), a proof-of-concept statement, and a forward-looking direction. The reader finishes knowing what was found, why it matters, and what the field should do next.

The structure is simple. Open with a stance on the primary finding. Highlight one or two results worth carrying forward. State what it means for practice or future knowledge. Close with one sentence pointing to the next step. Five to seven sentences is usually enough. This is not the place for new analysis or new literature — it's the place to land.

The failure mode is sagging. Researchers who've been careful and precise all the way through lose their nerve at the end. The conclusion becomes a pile of hedges — results should be interpreted cautiously, further research is needed, limitations apply. Or it speculates about mechanisms that belong in the Discussion. Or it just repeats the abstract, word for word, as if you've run out of things to say.

As a reader, that's deflating. A paper that built a careful argument across six sections deserves a landing that honours it. Learning how to write a conclusion is really just learning to hold your nerve for four more sentences.

Take a stand. Highlight what's most worth remembering. Say what comes next. Then stop.

Step 3: Keywords, title, abstract — written last, on purpose

These three get written last because by now the paper exists. You know exactly what it says, what it found, and what it contributes. Write a title or abstract before the paper is finished and you're describing a paper you haven't fully written — then revising that description every time the paper changes.

Write them in this order: keywords, then title, then abstract.

The logic is practical. Keywords are the terms your audience actually types when searching for research like yours. Your title should contain the most important of those terms — not because it sounds impressive, but because it's how people find you. The abstract is built last, from the skeleton you already have: the same five sentences (importance, background, gap, method, results) with one closing sentence added for the conclusion.

The temptation here is to get creative. A clever title feels like a reward for finishing. Resist it. In most academic fields, findability beats flair. A title that tells the reader exactly what you did — Group CBT with family and school involvement for adolescent anxiety: a randomised controlled trial — will be found and cited more reliably than one that gestures at something interesting without saying what's inside. Reviewers, editors, and search algorithms all respond to clarity.

Same with the abstract. It's not a teaser — it's a complete, compressed version of the paper. A reader who only reads the abstract should still know what you studied, what you found, and what it means. Abstracts that open with sweeping statements about the importance of the field, or bury the findings in the last sentence, aren't serving the reader. State the study, state the findings, state the implication. In that order.

And give keywords more thought than they usually get. Choose terms researchers in your field actually search for, not the vocabulary you happen to use. Check the keywords on the papers in your own background section — those are the terms that found you. They should be the terms that help others find your paper.

Clear beats clever. Write to be found.

Step 4: The final edit — one pass, at the very end

The final edit is the only stage where you're allowed to care about a comma.

Everything before this has been about argument, content, and structure. That's now settled. The final edit is for everything else: flow between paragraphs, repetitive language, consistent terminology, formatting conventions. One pass, at the end, after the content is done.

The order within the pass matters:

1. Flow. Do the paragraphs connect? Does each bridge cleanly to the next, or are there jumps the text doesn't support?
2. Clarity. Are there sentences doing too much? Phrases that show up three times in two paragraphs?
3. Terminology. Is the intervention called the same thing throughout, or has it drifted between "the program," "the treatment," "the intervention," and "the approach"?
4. Formatting. p-values, confidence intervals, effect sizes, table captions, reference style. Check everything against the journal's author guidelines in this pass — not before.

The most common mistake is doing all this as you write. The researcher who formats every p-value on the spot, polishes every sentence before moving on, and fixes every repetition the moment they notice it is a researcher who redoes all of that work when the content changes — and the content always changes. Polishing a sentence that gets cut next revision is time gone. Formatting p-values before the analysis is finalised is formatting you'll do twice.

The practical fix is a deferred list. During writing and feedback rounds, whenever someone points out a convention or you notice a rule you haven't been following, write it down. Don't fix it yet. Keep a running "final edit notes" document and add to it every time something catches your eye. When you reach the final edit, run through the list in one pass. Every issue, addressed once, at the right moment.

I spent real time early in my career formatting statistical output — p-values to the right decimals, confidence intervals in the correct notation, all of it done carefully — and then had to redo the lot when an analysis changed or a co-author pointed out the journal used a different convention. The deferred list would have saved every one of those hours.

One pass. Details last. Then submit.

The paper is finished

It's worth seeing how, because none of it depended on inspiration. You disarmed your reviewers by answering their three strongest objections before they could. You concluded with a stance instead of a shrug. You wrote the packaging last, when the paper actually existed to describe. And you saved every small detail for one deliberate pass at the very end, when nothing underneath it could change any more.

That's the workflow. It isn't glamorous, and that's its strength: every step is small, checkable, and repeatable — which means your next article will be faster than this one, and the one after that faster still.

Now submit it.

One question before you go: what reviewer objection are you most afraid of? Whatever it is — put it in your limitations section first, before they can. Hit reply and tell me which one it is. I read every reply.

📄 Free download: The 20-Point Final Polish Checklist — a one-page pre-submission sweep for formatting, flow, terminology, and journal conventions. Run it once, right before you submit.