How to Write a Discussion Section That Wows Reviewers

Reviewer 2 writes: "The results are interesting, but the discussion fails to contextualize them within existing literature." You've seen this feedback — maybe you've received it. It's the most common critique of discussion sections, and it points to a fundamental misunderstanding of what a discussion section is supposed to do.

Your discussion isn't a place to repeat your results in paragraph form. It's where you answer the question every reader has after seeing your data: "So what?"

We've reviewed thousands of discussion sections across disciplines. The ones that impress reviewers share a clear structure. The ones that get criticized share common mistakes. Here's how to write a discussion section that lands in the first category.

The discussion section structure that reviewers expect

Strong discussion sections follow a predictable arc. You don't have to label these segments — but you need to include all of them.

Opening: Restate your key finding in context. One paragraph. Start with your most important result and immediately connect it to the research question you posed in your introduction. Don't restate the statistical details — that's what the results section did. Instead, state the finding in plain language and explain what it means. "Our results demonstrate that X is associated with Y, supporting our primary hypothesis" is fine. Short. Direct.

Middle: Interpret findings in relation to existing work. This is the bulk of your discussion — typically 3–5 paragraphs. For each major finding, ask: Does this align with previous research? Contradict it? Extend it? This is where you bring back the literature you reviewed in your introduction and show how your findings fit into the bigger picture.

This interpretation layer is where most discussion sections fall short. We see two failure modes. First: researchers who just say "this aligns with Smith (2023)" without explaining why or what that alignment means. Second: researchers who discuss their findings in isolation, as if no one had ever studied the topic before.

The sweet spot is genuine intellectual engagement. "Our finding that X increases under condition Y aligns with Smith's (2023) threshold model but contradicts Lee's (2022) linear framework. The discrepancy likely reflects our use of ecological measures rather than laboratory tasks, suggesting the X-Y relationship may be context-dependent in ways previous work hasn't captured."

That's a discussion paragraph that moves knowledge forward.

Limitations: Be honest, specific, and constructive. Every study has limitations. Acknowledging them doesn't weaken your paper — it strengthens it by showing you understand the boundaries of your claims.

Name specific limitations: sample size constraints, measurement choices, generalizability boundaries, temporal limitations. For each one, briefly explain how it might affect interpretation. "Our sample was drawn exclusively from urban universities, which may limit generalizability to rural educational contexts where institutional resources differ substantially."

Don't turn your limitations section into an apology tour. State each limitation clearly, note its potential impact, and move on. Two to three paragraphs is usually sufficient.

Implications: What should the field do with your findings? Split this into theoretical implications (how your findings change our understanding) and practical implications (what practitioners, policymakers, or clinicians should do differently). Not every paper has both, but most have at least one.

Future directions: What comes next? Suggest 2–3 specific studies that would advance this line of research. "Future research should examine X" is too vague. "A longitudinal replication with annual measurement waves would test whether the X-Y relationship holds over time" — that's a suggestion someone could actually follow.

Connecting your discussion to your introduction

The best discussion sections create a satisfying loop. Your introduction posed a question. Your discussion answers it.

Reread your introduction before writing your discussion. Seriously. Open it in a split screen. Every claim you made about the gap in knowledge should be addressed — either confirmed, complicated, or left unresolved — in your discussion.

If your introduction said "no study has examined X in context Y," your discussion should explicitly state what you now know about X in context Y. If your introduction cited a debate between two theoretical frameworks, your discussion should explain which framework your findings support.

This echo between introduction and discussion gives your paper a coherence that reviewers notice. It signals that your study was designed with clear questions and that your discussion addresses them directly.

For guidance on structuring that introduction effectively, our guide to writing research paper introductions covers the funnel structure and gap statements in detail.

Common mistakes that weaken your discussion section

Repeating results instead of interpreting them. If a sentence in your discussion could be moved to your results section without changing anything, it doesn't belong in the discussion. "Participants in the treatment group scored 15% higher than control" is a result. "The 15% improvement in the treatment group exceeds the minimum clinically significant difference established by Park (2021), suggesting the intervention has practical — not just statistical — significance" is a discussion point.

Overclaiming. Your correlational study found an association between X and Y. Your discussion says "X causes Y." That leap will get your paper rejected. Match your discussion language to your methodology. Correlational designs get "is associated with," "predicts," and "relates to." Only experimental designs with proper controls earn causal language.

Ignoring unexpected findings. Your secondary hypothesis wasn't supported. Don't pretend it doesn't exist — discuss why. Null or unexpected results are often the most interesting part of a study. They suggest boundary conditions, measurement issues, or theoretical refinements that the field needs to hear about.

Listing limitations without analysis. "Our sample was small" isn't helpful. "Our sample of 45 participants may have been underpowered to detect the interaction effect, which would explain the non-significant moderation finding despite a trend in the expected direction" — that tells the reviewer you understand your data.

Skipping the practical implications. Especially in applied fields, reviewers want to know what practitioners should do with your findings. Even basic research has implications — they might just be for other researchers. Tell the reader who should care about your findings and what they should do about them.

Using AI to refine your discussion

The discussion section is deeply intellectual work — AI can't do the interpretation for you. But it can help you express your interpretations clearly.

We recommend running your discussion through our AI proofreader with a focus on three things. First, hedging consistency — make sure you're not overclaiming in one paragraph and underclaiming in the next. Second, paragraph length — discussion paragraphs that exceed 200 words usually contain two ideas that should be separated. Third, flow between paragraphs — the proofreader flags abrupt transitions where a connecting sentence would help.

The paraphrasing tool can help when you're struggling to express a complex interpretation clearly. Write your messy first attempt, then use the tool to generate alternative phrasings. Pick the one that captures your meaning most precisely — then edit it further in your own voice.

One specific technique we've found effective: write your discussion bullet points first. For each finding, write one sentence stating the result and one sentence stating what it means. Then expand each bullet into a paragraph, adding the connections to existing literature. This prevents the wandering that makes discussion sections lose focus.

AI Proofreader for Research Papers

Catch grammar errors, inconsistent hedging, and structural issues. Purpose-built for academic writing.

Frequently asked questions

Q: How long should a discussion section be?

For a standard journal paper, your discussion should be roughly 25–35% of the total word count — typically 1,500–2,500 words for a 7,000-word paper. The discussion is usually the longest section after the literature review. If yours is shorter than your results section, you probably haven't interpreted enough. If it's longer than your entire introduction and literature review combined, you've probably wandered off track.

Q: Should I discuss findings in the same order as the results section?

Start with your most important finding, regardless of where it appeared in the results. After that, you can follow the results order or organize by theme — whichever produces the most coherent narrative. The key is that the reader can clearly connect each discussion paragraph to the specific result it interprets. Using parallel subheadings between results and discussion sections can help with this mapping.

Q: How do I discuss non-significant results?

Don't hide them. Acknowledge the null finding, consider possible explanations (insufficient power, measurement issues, genuine absence of effect), and discuss what it means for the research question. A thoughtful discussion of null results shows more sophistication than ignoring them. If the null result contradicts previous findings, explore why. If it aligns with some prior work, note that. Non-significant doesn't mean unimportant.

Q: Can I introduce new references in the discussion?

Yes — and you should. Your discussion will often reference studies that didn't appear in your literature review, particularly when an unexpected finding connects to a body of literature you didn't anticipate needing. Introducing new references to explain, contextualize, or contrast with your findings is completely appropriate. Just don't turn the discussion into a second literature review.