Writing the Results Section of a Research Paper

The Results section has one job: report what was found, without arguing about what it means. That sounds simple, but it is the section where authors most often drift into the Discussion’s territory without noticing — explaining why a finding occurred, or how it compares to prior literature, before the reader has even finished seeing what was found. This guide covers how to structure a Results section, how to integrate figures and tables without duplicating them in prose, the statistical reporting conventions journals expect, and exactly where the line to Discussion sits. For sequencing Results within the rest of the drafting process, and for register/tense conventions that apply across the whole manuscript, see CASRAI’s How to Write a Research Paper guide, which this page extends. For the section that follows it, see Writing the Discussion Section of a Research Paper.

What the Results Section Is For

In the standard IMRaD (Introduction, Methods, Results, Discussion) structure used across most empirical scientific writing, Results and Discussion are deliberately separated so that readers and reviewers can evaluate the findings themselves before being told what to make of them. The Results section reports; the Discussion interprets. A Results section that starts explaining causes, comparing to prior work, or drawing conclusions is doing the Discussion’s job too early — and it is one of the most common structural weaknesses reviewers and editors flag, because it makes it harder to check whether an interpretation is actually supported by the data it claims to explain.

Concretely, a Results section needs to:

Report what was found, organized in a logical order — typically following the sequence of research questions, hypotheses, or the order procedures were introduced in Methods.
Present the data in appropriate form — prose for simple or singular findings, tables and figures for anything with more than a few data points worth comparing.
Report statistics precisely and consistently, following the target journal’s style (APA, AMA, or a discipline-specific convention).
Report negative and null results with the same rigor as positive ones, not omitted or downplayed because they are inconvenient.
Stay descriptive — state what a finding is, not why it happened or what it means for the field.

Structuring the Results Section

Most Results sections are organized around one of two logics, and the choice should be made deliberately rather than defaulting to whichever order the data happened to come out in:

By research question or hypothesis. If the Introduction set up two or three specific questions or hypotheses, presenting Results in that same order lets a reader track each question through to its answer without having to hold the whole study in mind at once. This is the more common approach in hypothesis-driven quantitative work.
By procedure, following Methods. If the study involved a sequence of stages (e.g., a screening step, then a main analysis, then a secondary or sensitivity analysis), reporting Results in that same sequence keeps the Methods-to-Results mapping obvious and easy to verify.

Whichever logic is used, subheadings that mirror the Methods section’s own subheadings (or the Introduction’s stated questions) make the paper easier to navigate and easier for a reviewer to check for completeness — a claim made in the Introduction that never resurfaces in Results is a gap reviewers are trained to look for. Within each subsection, a common and readable pattern is to state the finding first, in one sentence, then support it with the specific numbers, then point to the relevant table or figure. Leading with the number and only later revealing what it means forces the reader to hold an uninterpreted figure in mind until the sentence resolves.

Figure and Table Integration

Figures and tables in a Results section exist to carry information prose does the poorest job of — a distribution, a trend across conditions, a large comparison table with many rows. The prose and the display items should be complementary, not redundant:

Don’t restate a table in sentences. If a table already shows the mean, standard deviation, and sample size for six groups, the surrounding prose should point to the pattern the reader should notice (e.g., which group differed and by roughly how much), not recite all six rows in text. A common reviewer complaint is a Results section that duplicates every table value in prose, which adds length without adding information.
Every figure and table needs an explicit prose reference. (“As shown in Table 2…” / “Figure 3 shows…”) A table or figure that isn’t referenced anywhere in the text leaves the reader to guess why it’s there.
Captions should be self-contained. A reader skimming only the figures and tables (a common way papers are actually read) should be able to understand roughly what each one shows from its caption and axis/column labels alone, without needing the surrounding paragraph.
Order figures and tables in the sequence they’re first mentioned, and number them consistently with how the target journal expects (main text vs. supplementary).
Follow the journal’s technical figure requirements — resolution, file format, and image-integrity rules are a distinct submission concern from what belongs in a figure’s content. See CASRAI’s Scientific Figure Rules for Journal Submission guide for DPI, format, and integrity requirements.

Statistical Reporting Conventions

How statistics are reported in the Results section is one of the more heavily house-styled parts of a manuscript, and conventions differ by field and by journal style guide (APA, AMA, or a discipline-specific standard), but several practices are close to universal:

Report the test, the statistic, degrees of freedom (where applicable), and the p-value together, not the p-value alone — e.g., “t(48) = 2.31, p = .024” rather than just “p < .05”. A bare significance claim without the underlying statistic gives the reader nothing to check.
Report effect sizes alongside significance tests, not instead of them. A statistically significant result can correspond to a trivially small effect, and journals increasingly expect an effect-size measure (e.g., Cohen’s d, an odds ratio, r²) so the reader can judge practical, not just statistical, significance.
Report exact p-values rather than only threshold comparisons where the journal’s style permits it (e.g., “p = .032” rather than only “p < .05”), except at the extremes where a style guide specifies a minimum reportable value (e.g., “p < .001”).
Include confidence intervals where the study design supports them — a 95% CI communicates the precision of an estimate in a way a single point estimate or p-value does not.
Be consistent with decimal precision across the section — don’t report one statistic to four decimal places and another to one for no principled reason.
State sample sizes at each stage where attrition, exclusion, or missing data changed the analyzed N from the enrolled or collected N — a reader needs to know which N a given statistic is based on.

None of this replaces the target journal’s own instructions for authors, which take precedence whenever they specify a particular statistical reporting style. Where a study used a formal analysis or data plan, keeping the Results section’s statistical presentation consistent with how it was pre-specified (or clearly flagging where an analysis was exploratory rather than pre-specified) is expected practice, not an optional courtesy.

The Results-Discussion Boundary: Report, Don’t Interpret

The single most common structural error in a Results section is interpretation creeping in before the Discussion. In practice, this means avoiding certain kinds of sentences in Results, even when they feel natural to write in the moment:

Avoid explaining causes. “Group A improved more than Group B” is a Results statement. “Group A improved more than Group B because the intervention was more engaging” is a Discussion statement — the second half is an interpretation the data alone don’t establish.
Avoid comparing to prior literature. Noting how a finding relates to previous studies is core Discussion work; doing it in Results anticipates and partially duplicates that later section.
Avoid evaluative language about the finding’s importance. Words like “surprisingly,” “importantly,” or “as expected” smuggle in a judgment about what the result means, which belongs in Discussion.
Avoid stating implications. “This suggests that clinicians should reconsider X” is squarely Discussion territory, even if it feels like an obvious next sentence after reporting the finding it’s based on.

The practical test: could this sentence be written by someone who saw only the data and not the rest of the paper’s argument? If not, it likely belongs in Discussion rather than Results. Some journals and disciplines permit a combined “Results and Discussion” section (this is explicitly allowed under APA style for multi-study papers, for instance), in which case this separation applies less rigidly — but check the target journal’s own structure requirements before assuming a combined section is acceptable, since many biomedical and life-science journals require the two kept fully separate.

Reporting Negative and Null Results

A Results section should report what was found, including findings that don’t support the study’s hypothesis or that show no statistically detectable effect. Selectively omitting or downplaying null and negative results — sometimes called outcome reporting bias — distorts the evidence base the field relies on, and is a practice research-integrity and publication-ethics bodies have worked to counter, including through initiatives that specifically support publishing null results. A null result reported with the same statistical rigor as a positive one (the test used, the statistic, the confidence interval) is just as reportable as a positive one; what changes is only that the Discussion, not the Results, is where its implications get addressed.

Common Mistakes

Cherry-picking which comparisons to report. If a pre-specified analysis plan included several comparisons, all of them belong in Results, not only the significant ones.
Reporting methods details that belong in the Methods section. A Results section shouldn’t re-explain how a measure was administered or how a test was chosen — only what came out of applying it.
Inconsistent numbers between text, tables, and abstract. A statistic that differs between where it’s first reported and where it’s later cited (including in the abstract) is one of the most common reviewer catches and undermines confidence in the whole manuscript.
Vague quantification. “Many participants improved” without a number is weaker and less checkable than “68% of participants (34/50) improved.”
Overlong prose restating tabular data. See the figure/table integration guidance above — if a table already shows it clearly, the prose job is to point to the pattern, not repeat every value.

Frequently Asked Questions

How is the Results section different from the Discussion section?

Results reports what was found — the data, statistics, and outcomes of the study’s analyses — without explaining why those findings occurred or what they mean. Discussion interprets those findings: it explains their significance, compares them to prior work, states limitations, and draws conclusions. See CASRAI’s Writing the Discussion Section guide for the interpretive half of this pair.

Should I report every statistical test I ran, or only the significant ones?

Every test that was part of the study’s pre-specified analysis plan should be reported, whether or not it reached statistical significance. Reporting only significant results while omitting non-significant ones from the same analysis plan misrepresents the evidence and is a recognized form of selective reporting.

Can I combine Results and Discussion into one section?

Some journals and style guides permit it — APA style, for example, explicitly allows a combined “Results and Discussion” section in multi-study papers, provided a comprehensive general Discussion still closes the article. Many biomedical and life-science journals expect them kept separate. Check the target journal’s instructions for authors before combining them.

How long should a Results section be?

There is no fixed word count; length depends on how many analyses the study involved and how much can be offloaded to tables and figures rather than prose. A well-integrated Results section is usually shorter than a first draft, because moving detailed numbers into tables and keeping prose focused on the pattern each table or figure shows trims length without losing information.

Do I need to report confidence intervals as well as p-values?

Where the study design supports them, yes — confidence intervals communicate the precision of an estimate in a way a p-value alone does not, and many journals now expect them alongside significance tests as standard practice.

Where should I discuss a result’s limitations?

In the Discussion section, not Results. A Results section can note a factual limitation of the data itself (e.g., missing data at a specific stage that changed the analyzed sample size), but the interpretive question of how a limitation affects what the finding means belongs in Discussion.

For section-by-section coverage of the rest of the manuscript, see CASRAI’s How to Write a Research Paper guide and the Research Paper Format guide for formatting conventions across the whole document.