Writing Results and Discussion Sections That Work

The results and discussion are where a paper earns its conclusions. The results show what you found, without spin; the discussion explains what the findings mean, how they fit with earlier work and how far they can be trusted. Most reviewer criticism lands here: interpretation creeping into the results, discussions that merely repeat numbers, claims that outrun the data, and limitations squeezed into one sentence. This guide covers structure, reporting, interpretation, limitations, implications and language, with examples for quantitative and qualitative studies.

What the results and discussion each do

The two sections answer different questions. The results answer “What did we find?” They present data relevant to each research question or hypothesis, in a logical order, with the statistics or qualitative evidence needed to judge them. They do not explain why the findings occurred or what they mean for practice.

The discussion answers “So what?” It states whether the questions were answered, interprets the findings, compares them with earlier research, acknowledges limitations and draws implications. It should not introduce new results, and it should not simply restate the numbers.

Keeping this division clear helps reviewers. If a statistic appears for the first time in the discussion, they will ask why it is missing from the results; if the results are full of “this suggests” and “interestingly”, they will ask whether you can separate evidence from opinion. Every analysis reported should also be described in the methods; see Writing a Reproducible Methods Section to make sure the two line up.

Separate or combined results and discussion?

Many journals require separate sections; others allow or prefer a combined “Results and Discussion”. Check the author guidelines before deciding.

If you combine them, follow a fixed pattern for each finding: report it, then interpret it, then compare it with the literature. End with a short synthesis, a limitations subsection and implications, so that the overall meaning does not disappear among individual findings.

Organising the results section

Structure the results around your research questions or hypotheses, not around the order in which you ran the analyses.

1. Start with the sample. Report response rates, participant or sample characteristics and, for trials, the flow of participants. CONSORT and STROBE both ask for this information.
2. Then answer each question in turn, using subheadings that echo the research questions or hypotheses from the introduction.
3. Put primary outcomes before secondary ones, and pre-specified analyses before exploratory ones, labelling the latter clearly.
4. Report all planned analyses, including those that did not support your hypothesis. Selective reporting is a recognised form of bias.

Open each subsection with a sentence that tells the reader what is being reported (“Table 2 shows attendance by type of internet access”), then give the key numbers. Keep raw data, long robustness checks and extra tables for supplementary material, and say where readers can find them.

Reporting findings objectively

Objective reporting means giving readers enough information to judge a finding themselves.

• Report effect sizes with confidence intervals, not only p-values. A difference can be statistically significant yet too small to matter.
• Give exact p-values (p = .03 rather than p < .05), unless very small (p < .001).
• Include denominators and units: “42 of 120 students (35%)”, not “35% of students”.
• Use neutral verbs: showed, was associated with, did not differ. Save “because” and “suggests” for the discussion.
• Report null results plainly. “Attendance did not differ by gender (difference 1.2 classes, 95% CI −0.8 to 3.2)” is a finding, not a failure.

The American Statistical Association’s 2016 statement on p-values makes the same point: a p-value alone does not measure the size or importance of an effect.

Making text, tables and figures work together

Text, tables and figures should complement each other, not repeat each other.

• Tables suit exact values and many variables: descriptive statistics, regression coefficients, item-level scores.
• Figures suit patterns, trends, comparisons and distributions.
• Text highlights what matters most: the direction and size of the main effects, the key comparison and anything unexpected.

Refer to every table and figure in the text, in numerical order, and tell the reader what to notice (“Figure 2 shows that the gap widened after the second term”). Do not repeat every value in a table in the prose. Each table and figure should be understandable on its own, with a clear title, labelled axes or columns, units and notes defining abbreviations and statistical symbols.

Avoid presenting the same data twice, as both a table and a chart. For formatting details, see Preparing Figures and Tables for Journals and the APA Style guidance on tables and figures.

Reporting qualitative findings

In qualitative studies the results usually present themes or categories, each supported by evidence from the data.

• Start with an overview of the themes, perhaps in a short table, before describing each one.
• Define each theme in a sentence, then illustrate it with short, well-chosen quotations.
• Identify quotations with participant codes (for example P7, teacher, rural school) rather than names, protecting confidentiality.
• Show variation, including views that differ from the dominant pattern.
• Be careful with counting words. “Most” and “a few” should reflect the data, and some journals prefer you avoid frequency claims altogether.

Quotations are evidence, not decoration: introduce each one and say what it shows. Many qualitative papers combine results and discussion. Reporting guidelines such as COREQ and SRQR, listed on the EQUATOR Network, set out what readers need to assess qualitative findings.

Opening the discussion and interpreting findings

Begin the discussion with a short paragraph stating the main findings in words, as direct answers to your research questions. Readers should not have to reconstruct them from statistics.

Then interpret. For each important finding, ask:

• What is the most plausible explanation? Link it to theory or to known mechanisms.
• What else could explain it? Confounding, measurement error, sampling or chance. Address each seriously.
• How large and how certain is it? Use the confidence interval to say whether the effect is practically meaningful and how precise the estimate is.
• Does it fit the design? Cross-sectional and observational data support “associated with”, not “caused”.

Discuss unexpected and null findings too; they are often the most informative part of a paper. Organise the discussion by theme or research question rather than repeating the results section paragraph by paragraph.

Comparing your findings with the literature

Comparison shows where your study sits in existing knowledge. It is not a second literature review: cite only studies that help explain or contrast your findings, many of which should already appear in your introduction.

• When findings agree, say what your study adds: a new population, a larger sample, a stronger design or a more precise estimate.
• When findings differ, look for reasons: different populations, settings, time periods, measures or analytical methods. Treat differences as informative, not embarrassing.
• When no comparable study exists, compare with the closest evidence and say so honestly.

Close the loop with the introduction: the gap you described there should be visibly narrowed here.

Writing limitations honestly

Every study has limitations, and reviewers trust authors who identify them before they do. A useful limitations paragraph is specific and explains consequences, not just a generic list.

For each important limitation, state three things: what the limitation is, how it may have affected the results (direction and size, if possible), and what you did to reduce it, such as a sensitivity analysis.

Common categories include sampling and generalisability, measurement, design (for example, no causal inference from cross-sectional data), missing data and unmeasured confounders. Avoid two extremes: a single line such as “the sample size was small” with no explanation, or a list so long that it undermines the whole study. Many authors place limitations near the end of the discussion, before implications; some journals ask for a dedicated subheading.

Implications and future research

Implications tell readers what to do with your findings. Make them specific and keep them within what the data support.

• Theoretical: what the findings mean for existing theories or models.
• Practical: what practitioners, managers, teachers or clinicians might do differently.
• Policy: where findings are relevant to regulation or funding, phrased cautiously when evidence comes from a single study.

Recommendations for future research should be concrete. “More research is needed” helps no one; “a longitudinal study following students over a full academic year would show whether the attendance gap affects examination results” gives other researchers a starting point.

End with a short conclusion, either as the final discussion paragraph or as a separate section, depending on the journal. Restate the main answer and its significance without new data or citations. Before the reference list, many journals now expect a Data Availability Statement; JMVRA, for instance, asks for one in that position. See Data Availability Statements: 6 Templates and Tips.

Tense, hedging and common mistakes

Hedge in proportion to your evidence: “suggests”, “may” and “is consistent with” for observational or exploratory findings; firmer wording only for strong, replicated or experimental results. Over-hedging (“it might possibly perhaps indicate”) is as unhelpful as overclaiming.

Common mistakes reviewers flag:

• Interpretation or citations inside the results section.
• A discussion that repeats the results without explaining them.
• New data appearing for the first time in the discussion.
• Causal language for correlational findings, or generalising beyond the sample.
• Ignoring null or contradictory results.
• Limitations that are missing, generic or buried.

Under double-blind review, such as JMVRA’s peer review, reviewers judge these sections on the evidence alone. For handling reviewer comments afterwards, see Revising a Paper After Review.

Quick Checklist

1. Results by questionEach RQ or hypothesis answered in the same order
2. Full statisticsEffect sizes, intervals, exact p-values, denominators
3. No double reportingText highlights; tables and figures hold the detail
4. Answer firstDiscussion opens with the main finding in words
5. Specific limitationsEach states its likely effect on the results
6. Bounded implicationsClaims stay within the design and the sample

Five habits for clearer results and discussion

These habits help most when you revise, after a full draft of both sections exists.

• Draft results from your tables. Finalise tables and figures first, then write the text around what each one shows.
• Highlight interpretive words. Search the results for words such as suggests, because, interestingly and clearly, and move those sentences to the discussion.
• Write the first discussion paragraph last. Once the discussion is complete, rewrite its opening so it states the main answers precisely.
• Test every claim against the design. For each sentence in the implications, ask whether your design and sample genuinely support it.
• Read the introduction and discussion together. The gap, the questions and the contribution should match across both sections.

Let the results show what you found and the discussion explain what it means, and never let either do the other's job.

Questions Researchers Often Ask

In Summary

Good results and discussion sections divide the work cleanly. The results present findings in the order of your research questions, with effect sizes, intervals and well-designed tables and figures, and without interpretation. The discussion opens with the answer, explains it, tests it against alternative explanations and earlier studies, states limitations with their likely consequences, and draws implications no stronger than the design allows. Write in the past tense for your findings and the present for their meaning, hedge in proportion to the evidence, and check that the discussion fulfils every promise made in the introduction.

References and Sources

• EQUATOR Network — Reporting guidelines library
• ICMJE — Recommendations for the Conduct, Reporting, Editing and Publication of Scholarly Work
• APA Style — Tables and figures
• STROBE — Reporting observational studies