The results section reports what you found — objectively, in past tense, with no interpretation. Present your findings in the same order as your methods, lead with your primary outcome, report effect sizes and confidence intervals alongside p-values, and use tables and figures for data while reserving the text for what matters most. Interpretation belongs in the discussion, not here.
Most researchers find the results section the easiest to draft and the hardest to get right. The data already exist, so the writing feels mechanical — and that is exactly the trap. Reviewers routinely flag results sections for selective reporting, duplicated data, missing statistics, and interpretation that has crept in from the discussion. This guide covers what to include, how to decide between text, table, and figure, how to report statistics correctly, and the errors that cost authors revisions.
What Is the Results Section?
The results section presents the findings of your analysis without explaining or evaluating them. It answers one question: what happened?
The discipline required here is unusual in academic writing. You are reporting, not arguing. Your job is to give readers the evidence in a form they can assess independently — so that when you interpret it in the discussion, they can judge whether your interpretation is fair.
The line between results and discussion
| Sentence | Belongs in |
|---|---|
| “Adherence was 78.2% versus 61.4% (p < 0.001).” | Results |
| “Adherence was higher, likely because of more frequent contact.” | Discussion |
| “Systolic blood pressure fell by 8.3 mmHg.” | Results |
| “This reduction is clinically meaningful.” | Discussion |
| “The effect was consistent across age groups.” | Results |
| “The consistency suggests the mechanism is not age-dependent.” | Discussion |
A practical test: if a sentence contains because, suggests, importantly, or surprisingly, it has crossed into interpretation. Move it.
Note the final pair, which is where most authors go wrong. Describing a pattern in your data is still reporting. Saying the effect was consistent across age groups is an observation about what the data show. Explaining what that consistency implies is interpretation. The boundary is not between simple and complex statements — it is between describing and explaining.
Some journals permit a combined “Results and Discussion” section, which is common in chemistry and materials science. Check your target journal’s recent articles before merging them.
What to Include
| Component | Include? | Notes |
|---|---|---|
| Participant or sample flow | Yes | Numbers screened, included, excluded, analyzed |
| Baseline characteristics | Yes | Usually Table 1 |
| Primary outcome | Yes — first | The result your study was designed to answer |
| Secondary outcomes | Yes | After the primary |
| Subgroup or sensitivity analyses | Yes | Label clearly as pre-specified or exploratory |
| Null or negative findings | Yes | Omitting them is selective reporting |
| Interpretation | No | Discussion |
| Methods detail | No | Methods |
| New analyses not in methods | No | Add to methods, or label as post hoc |
Step-by-Step: Writing the Results Section
Step 1 — Mirror the order of your methods
If your methods described recruitment, then the primary outcome, then two secondary outcomes, your results should follow the same sequence. Reviewers read the two sections against each other, and a mismatch immediately raises the question of whether something was reported selectively.
Step 2 — Open with participant or sample flow
Tell the reader what data the analysis actually rests on before you report any finding.
“Of 612 patients screened, 480 met the eligibility criteria and were randomized. Twenty-nine participants (6.0%) were lost to follow-up, leaving 451 in the final analysis. Attrition did not differ significantly between arms.”
Clinical trials generally require a flow diagram following CONSORT; systematic reviews follow PRISMA. Check which reporting guideline applies to your study design.
Step 3 — Lead with your primary outcome
Do not build suspense. The finding your study was designed to test comes first, reported with its effect size, precision, and significance.
Weak: “There was a statistically significant difference between the groups (p < 0.001).” Strong: “Adherence was higher in the intervention group than in controls (78.2% vs 61.4%; difference 16.8 percentage points, 95% CI 9.7–23.9; p < 0.001).”
The weak version tells you a test was passed. The strong version tells you what was found, how large it was, and how precisely it was estimated — which is what a reader needs to judge the result rather than trust it.
Step 4 — Report secondary outcomes, then any additional analyses
Keep the hierarchy visible. Label exploratory or post hoc analyses explicitly; presenting them with the same weight as pre-specified outcomes is a reporting failure reviewers watch for.
“Mean systolic blood pressure decreased by 8.3 mmHg (SD 4.1) in the intervention group compared with 2.1 mmHg (SD 3.8) in controls (p = 0.002). In a pre-specified subgroup analysis, the effect was consistent across age groups.”
Step 5 — Report null and negative findings honestly
Findings that did not reach significance still belong in your paper. Report them plainly, without apology and without spin.
Avoid: “Although not statistically significant, quality-of-life scores showed a promising trend towards improvement (p = 0.09).” Better: “Quality-of-life scores did not differ significantly between groups (mean difference 1.4 points, 95% CI −0.3 to 3.1; p = 0.09).”
“Trending towards significance” is one of the phrases reviewers most reliably object to. A result is either statistically significant at your pre-specified threshold or it is not.
Step 6 — Decide what goes in text, table, or figure
This decision causes more revision requests than almost anything else in the results section, because authors duplicate the same data in two places.
| Use | When | Example |
|---|---|---|
| Text | Two or three numbers; the headline finding | The primary outcome |
| Table | Many precise values; readers need exact numbers | Baseline characteristics, full outcome data |
| Figure | Patterns, trends, distributions, or comparisons over time | Blood pressure trajectory across 12 months |
The rule that matters: never present the same data in two formats. If a value appears in a table, the text should highlight its significance rather than repeat it — “adherence differed significantly between arms (Table 2)” rather than restating every figure.
Step 7 — Write text that guides rather than duplicates
The text is not a narration of your tables. Its job is to point the reader to what matters.
Weak: “Table 2 shows the results. Adherence was 78.2% in the intervention group. Adherence was 61.4% in the control group. Blood pressure decreased by 8.3 mmHg in the intervention group and 2.1 mmHg in the control group.” Strong: “The intervention improved both adherence and blood pressure control, with the adherence difference exceeding the pre-specified 10 percentage point threshold for clinical relevance (Table 2).”
Step 8 — Format tables and figures to journal standards
Every table and figure should be interpretable on its own: a self-explanatory caption, defined abbreviations, stated units, and a stated sample size. Number them in the order they are first cited in the text. Confirm your journal’s requirements for file format, resolution, and colour before submission — our pre-submission formatting checklist covers the common breaches.
How to Report Statistics Correctly
Incomplete statistical reporting is among the most frequent reasons for revision requests. Use this as a reference.
| Report | Correct form | Common error |
|---|---|---|
| Exact p-values | p = 0.03, p = 0.002 | p < 0.05 for everything |
| Very small p-values | p < 0.001 | p = 0.0000 |
| Effect size | Mean difference, odds ratio, Cohen’s d | Significance with no magnitude |
| Precision | 95% CI 9.7–23.9 | Point estimate alone |
| Variability | Mean (SD) or median (IQR) | Mean with no spread |
| Sample size | n for every group and analysis | n stated only once at the start |
| Decimal places | Consistent, matched to precision | 78.24316% |
| Percentages | With the raw count: 78.2% (n = 188) | Percentage alone in small samples |
Two principles underpin the table. First, an effect size without precision is incomplete, and significance without an effect size is close to meaningless — a p-value tells you whether an effect was detected, not whether it matters. Second, report the statistic your analysis actually produced; do not convert results into a form that looks stronger.
One further distinction is worth guarding carefully: in a results section, “significant” should mean statistically significant and nothing more. A trivial difference can reach significance in a large sample, and a meaningful one can miss it in a small sample. Whether a finding actually matters is a judgment, and judgments belong in the discussion.
Tense and Voice
- Past tense for what you found: “adherence was higher”, “scores decreased”.
- Present tense for what tables and figures do: “Table 2 shows”, “Figure 1 presents”.
- Either voice is acceptable in most journals. “We measured” is now widely accepted and often clearer than the passive alternative, but follow your journal’s convention.
How Long Should the Results Section Be?
Typically 500–1,000 words for a standard research article, or 15–20% of the manuscript body — usually shorter than the methods and considerably shorter than the discussion.
| Signal | What it usually means |
|---|---|
| Results longer than the discussion | Data dumping, or duplicating tables in the text |
| Results under 300 words with many outcomes | Under-reporting; reviewers will ask for more |
| Frequent use of “interestingly” or “notably” | Interpretation has crept in from the discussion |
Results Section Checklist
- Follows the same order as the methods
- Opens with participant or sample flow
- Primary outcome reported first
- Effect sizes and confidence intervals reported, not just p-values
- Exact p-values given
- Sample sizes stated for every analysis
- Null and negative findings included
- Exploratory analyses labelled as such
- No data duplicated between text and tables
- Tables and figures interpretable standalone
- Past tense throughout
- No interpretation, speculation, or new methods
Common Mistakes to Avoid
- Interpreting your results. The most common error by a wide margin. Save meaning for the discussion.
- Duplicating data across text and tables. Reviewers ask for one or the other, not both.
- Reporting only significant findings. Selective reporting is a research integrity issue, not a style preference.
- Giving p-values without effect sizes. Significance without magnitude tells readers very little.
- “Trending towards significance.” A frequent and easily avoided reviewer objection.
- Narrating tables sentence by sentence. The text should guide attention, not repeat data.
- Introducing analyses not described in the methods. Either add them to the methods or label them post hoc.
- Inconsistent decimal places or units. Small inconsistencies suggest carelessness with the data itself.
Careless reporting here contributes to a meaningful share of desk rejections — see why papers get desk rejected for the wider picture.
Frequently Asked Questions
What is the difference between the results and discussion sections? The results section reports what you found, objectively and without explanation. The discussion interprets those findings, compares them with prior research, and explains what they mean. If a sentence explains why a result occurred, it belongs in the discussion.
Should I include statistical tests in the results section? Yes. Report the test outcome with an effect size, a measure of precision such as a 95% confidence interval, and an exact p-value. The test method itself belongs in your methods section.
Do I have to report non-significant results? Yes. Omitting findings that did not reach significance is selective reporting and a recognized research integrity problem. Report them plainly, without describing them as trending towards significance.
Should data appear in both a table and the text? No. Present each piece of data once. Use tables for precise values and reserve the text for highlighting what matters, referring the reader to the table rather than repeating its contents.
How long should a results section be? Usually 500–1,000 words, or roughly 15–20% of the manuscript body. It is normally shorter than both the methods and the discussion.
What tense should the results section use? Past tense for your findings, since the analysis is complete. Use the present tense only when referring to what a table or figure shows.
Can the results and discussion be combined? Some journals allow or prefer a combined section, particularly in chemistry, materials science, and engineering. Check your target journal’s author guidelines and recent published articles before deciding.
Conclusion
A strong results section is disciplined rather than clever. Follow the order of your methods, lead with the primary outcome, report effect sizes and confidence intervals rather than significance alone, include the findings that did not go your way, and present each piece of data exactly once. Keep interpretation out entirely — and the discussion that follows becomes far more persuasive, because readers have already seen the evidence on its own terms.
Want your data reported the way reviewers expect?
Our subject-matter editors check statistical reporting, table and figure consistency, and journal compliance across the full manuscript. Start with a free 500-word sample edit, no obligation.
Request your free sample edit → or explore our scientific editing services and journal formatting support.
