How to Write the Results Section of a Research Paper

The results section of a research paper presents what the study found. It reports the evidence produced by the research design, explains how the data were analysed at the level needed to understand the findings, and directs readers to the tables, figures, quotations or other materials that document the results. Its central responsibility is accurate reporting. It does not replace the discussion section, introduce a new literature review or argue that the findings prove more than the data can support.

When learning how to write results section of a research paper, begin by separating three connected but different tasks: collecting or assembling the evidence, analysing it according to the stated method, and reporting the output in a form that answers the research questions. The results section covers the third task. It should make the findings easy to locate, understand and verify.

In a quantitative paper, results may include participant flow, sample characteristics, descriptive statistics, inferential tests, effect estimates, confidence intervals and model diagnostics. In a qualitative paper, they may include themes, categories, patterns, participant quotations and evidence of variation across cases. A mixed-methods paper reports the findings from each strand and explains where the two forms of evidence are brought together.

The section is not simply a place to paste software output or reproduce every observation. It is a selective, structured account of the findings that matter to the research question. Selection must be transparent: report outcomes specified in the protocol or analysis plan, explain justified deviations and avoid suppressing results because they are unexpected or statistically non-significant.

In a conventional empirical research paper, the results section follows the methodology or materials and methods section and precedes the discussion. A common sequence is Introduction, Literature Review or Background where required, Methods, Results, Discussion and Conclusion. The exact structure varies. Some journals combine Results and Discussion; some dissertations separate findings and discussion into distinct chapters; and some qualitative traditions use thematic findings as a chapter with a discipline-specific structure.

The results section must align with the research questions, objectives, hypotheses and analysis plan introduced earlier. If the paper asks whether two variables are associated, the findings should report the relevant association estimate and uncertainty. If it asks how participants experience a service, the findings should present the themes and supporting evidence that answer that experiential question. If it evaluates an intervention, the results should report outcomes at the specified time points and explain the participant numbers contributing to each analysis.

Use the same terminology for constructs, groups, variables and outcomes throughout the paper. If the methods section calls a group “usual care,” do not rename it “control group” in a way that could imply a different design. Define abbreviations before using them and retain consistent labels across prose, tables, figures and appendices.

Before drafting, create a simple alignment map: research question, outcome or theme, analysis, finding, and location in the results section. This prevents a common problem in which the section contains many statistics or quotations but does not clearly answer the stated question.

Writing should begin only after the analysis dataset and outputs have been checked. Confirm that the dataset corresponds to the approved study version, variables have the correct labels and units, exclusions are documented, missing values are handled as planned and calculations can be reproduced. For qualitative work, confirm that transcripts, field notes, documents and coding records correspond to the material described in the methods section.

In quantitative research, inspect frequencies, ranges, impossible values, duplicate records, coding errors and missingness. Check that categorical values use consistent labels and that continuous variables have plausible units. If data cleaning changes the sample, document the reason and retain an audit trail. Do not remove observations simply because they weaken a hypothesis or make a result less convenient.

For qualitative research, check transcript accuracy, anonymisation, coding consistency and the connection between each proposed theme and the source material. If more than one researcher coded the data, report the relevant process where required by the chosen methodology. Preserve contextual details that are necessary to understand an excerpt without revealing participant identities.

Verify analysis scripts, formulas and software settings. Re-run key analyses where feasible and compare the results with the tables you plan to report. Keep a record of software name and version if the discipline, journal or assignment requires it. Never invent, alter or selectively omit data to create a more persuasive narrative.

The opening paragraph should orient readers without repeating the entire methodology. Depending on the study, it may state the number of participants included in the analysis, the order in which findings are presented, the primary outcome or the broad structure of the themes. It can also identify where detailed participant characteristics or preliminary checks appear.

A useful opening is specific. For example: “Of the 184 participants enrolled, 171 completed the 12-week assessment and were included in the primary outcome analysis. Findings are presented in the order of the three study objectives, beginning with changes in the primary outcome.” The numbers in an actual paper must, of course, match the verified study records.

Avoid opening with generic statements such as “This section presents the results of the research.” That sentence adds little. Also avoid announcing that the hypothesis was “proven.” Research findings provide evidence in support of, or inconsistent with, a hypothesis; they do not establish certainty through a single test.

If the results section is short, an opening sentence may be sufficient. If it is a long dissertation chapter, a brief roadmap can help readers navigate the sections.

Descriptive statistics summarise the characteristics of the sample and the distribution of measured variables. Common summaries include frequencies and percentages for categorical variables, and means with standard deviations or medians with interquartile ranges for continuous variables. Choose the summary that fits the variable’s scale and distribution, not simply the one that looks familiar.

A mean and standard deviation are often suitable for approximately symmetric continuous data. A median and interquartile range may be more informative for skewed data or ordinal measures. Report minimum and maximum values when they help readers understand the range. For categorical variables, provide counts and percentages and make clear whether percentages are based on the full sample or valid responses only.

Include units. A mean age of 42.6 is incomplete unless readers know the unit is years. Likewise, a blood pressure, temperature, distance or concentration should be reported with the appropriate measurement unit and consistent precision. Define scale scores and explain whether higher values represent more or less of the construct.

Avoid reporting every demographic variable in prose if a table presents them clearly. The text should identify the most relevant sample features and point to the table. Do not interpret baseline differences as evidence of an intervention effect unless the design and analysis justify that conclusion.

Report findings that do not support the hypothesis or fail to reach a conventional significance threshold. A non-significant result is not proof that no relationship or effect exists. It may reflect a small effect, imprecise measurement, limited sample size, high variability, design constraints or a genuinely weak association. The results section should state what the analysis found without turning a failure to reject a null hypothesis into proof of equivalence.

Use estimates and confidence intervals to show the direction, magnitude and precision of the result. If the study was designed as an equivalence or non-inferiority study, report the prespecified margin and the analysis appropriate to that design. Do not reinterpret an ordinary null-hypothesis test as an equivalence test after seeing the data.

Unexpected findings should be reported when they are relevant and credible. Identify whether they were prespecified, secondary or exploratory. Avoid hiding inconvenient results, selectively reporting favourable outcomes or changing the primary outcome after analysis without transparent explanation.

Keep explanations for unexpected results mainly in the discussion. In the results section, report the observation, relevant estimate and analysis. Later, consider plausible explanations and whether they require further study.

Participant quotations give readers direct access to the material underpinning a qualitative interpretation. Select excerpts that are relevant, sufficiently contextualised and representative of the point being made. A long quotation is not automatically stronger evidence. Often a concise extract followed by analysis is clearer.

Introduce each quotation so readers know who is speaking in the study’s anonymised labelling system and why the extract is relevant. After the quotation, explain the feature it illustrates. Do not assume the quotation speaks for itself. Where appropriate, include contrasting excerpts to show that participants had different experiences or that a theme contained tension.

Use a consistent transcription style. If pauses, emphasis, non-verbal behaviour or researcher notes matter to the analysis, explain the notation in the methods section or a note. Avoid adding punctuation or grammatical corrections that change the participant’s meaning. If a translation is used, explain the translation process where it affects interpretation.

Check the consent form, ethics approval and institutional rules before publishing excerpts. Remove names, locations, employers and distinctive details that could identify a participant. Anonymisation should protect privacy without removing the context needed to understand the finding.

Tables are useful when readers need to compare multiple values, groups, time points, categories or model estimates. A good table is understandable without requiring readers to search repeatedly through the main text. Give it a clear number and descriptive title, label rows and columns, include units, define abbreviations and provide notes for statistical symbols or analytical details.

Use a table when exact values matter or when several related results need to be compared. Do not create a table for one or two numbers that can be stated clearly in a sentence. Avoid duplicating every table value in prose. Instead, describe the most important pattern and direct readers to the table for detail.

Keep decimal places consistent within comparable columns. Align numbers by decimal point where the software or document format allows. Use a consistent number of significant digits and avoid false precision. Report denominators for percentages and clarify whether missing observations were excluded from a calculation.

Every table should be mentioned in the text in numerical order. Check that the table agrees with the narrative, abstract and supplementary files. A table that contradicts the prose undermines confidence in the whole paper.

A strong results paragraph usually has one main purpose. Begin with a sentence that identifies the question, outcome, theme or comparison. Follow with the key finding and the evidence needed to evaluate it. Then point to a table, figure or quotation if it contains supporting detail. Keep interpretation proportionate to the results-section role.

For quantitative writing, a useful sequence is: identify the outcome; state the comparison or association; report the estimate and uncertainty; provide the test result where required; and refer to the relevant table. For qualitative writing, introduce the theme; explain its meaning; provide a data extract; and analyse how the extract supports the theme.

Use precise verbs. “Increased,” “decreased,” “was associated with,” “differed,” “clustered,” “participants described” and “the model estimated” communicate more than vague phrases such as “had an impact” or “showed a lot.” Use causal language only when the design and analysis support causal inference.

Keep sentences manageable. Place the finding near the beginning rather than burying it after a long methodological reminder. Define abbreviations and avoid unexplained statistical shorthand. If a sentence contains several estimates, consider whether a table would make the comparison easier to follow.

Statistical wording affects how readers interpret a finding. Use “associated with” for an observed relationship unless the study design and analysis justify causal wording. Avoid saying that a variable “caused” an outcome in a cross-sectional observational study. A regression coefficient does not by itself prove causation, and adjustment for measured covariates does not eliminate all confounding.

Use “statistically significant” only when the relevant test and threshold justify it. Do not describe a result as important solely because p is below .05. Report the effect estimate and uncertainty, and distinguish statistical significance from clinical or practical significance. If a confidence interval includes the null value, describe the estimate and interval accurately rather than claiming that the variables are unrelated.

Be careful with terms such as “trend.” A result that does not meet a prespecified threshold should not automatically be described as a trend toward significance. State the estimate, interval and p-value, then let readers evaluate the evidence. Likewise, do not say that a hypothesis was proven or disproven by one analysis.

Use terminology consistently: participants, respondents, patients, records or cases should reflect the actual study population. Distinguish sample size from the number of observations, and distinguish standard deviation from standard error. These measures answer different questions.

Missing data can affect sample size, precision and the interpretation of findings. State how much data were missing for important variables, how missingness was handled and how the number of observations differs across analyses. If cases were excluded, give the criteria and counts where relevant. The methods section should explain the procedure; the results section should report what happened in the analysed data.

Do not describe a complete-case analysis as if all enrolled participants contributed to every outcome. State the denominator for each analysis or make the sample sizes clear in a table. If imputation was used, report the number of imputed datasets and the relevant procedure as required by the study design and reporting guideline.

For qualitative studies, explain whether any interviews or documents were incomplete or excluded and whether that affected the dataset. Do not silently remove an account because it contradicts the emerging themes.

Transparent reporting helps readers assess potential bias. It also makes the paper reproducible because readers can understand which observations contributed to each reported finding.

Survey and observational research commonly reports sample characteristics, response or participation information, outcome distributions and associations between variables. Explain the denominator for each percentage and whether weights, clustering or stratification were applied. If the sample was not probability-based, avoid implying that it represents a broader population without justification.

For cross-sectional studies, report associations without implying temporal order or causation. For cohort studies, state the follow-up period and the number contributing to outcome analyses. For case-control studies, report the relevant case and control counts and effect estimates appropriate to the design. Use STROBE as a reporting aid for observational research.

Describe the actual measure used. If a construct such as job satisfaction, anxiety or health literacy was measured with a scale, report the score direction and relevant summary. If the measure was developed for the study, explain its scoring and limitations in the methods and report its observed performance where appropriate.

Be cautious when interpreting self-reported information. Recall, social desirability and non-response may affect survey findings. These limitations are usually discussed later, but the results should accurately describe what respondents reported.

Laboratory results should report observations and measurements in a reproducible and disciplined way. Include the relevant units, number of replicates, uncertainty or variability measure and conditions needed to interpret the values. Distinguish raw observations from calculated quantities and explain any transformation or normalisation in the methods.

Use tables for exact measurements and figures for patterns or relationships. Label axes, define abbreviations and identify the control or reference condition. If an experiment produced an unexpected observation, report it honestly and distinguish it from a confirmed measurement or a speculative explanation.

Do not alter images, remove inconvenient data points or adjust values to match an expected result. If an instrument malfunction, contamination or protocol deviation affected the experiment, document it and explain its impact. Exclusions should be justified by a predefined rule or a defensible technical reason, not by whether the result supports the hypothesis.

In engineering and physical sciences, include the precision and units appropriate to the measurement system. In biological sciences, identify sample sizes and biological versus technical replicates. The terminology must match the actual experimental design.

This workflow works for a short empirical assignment and can be expanded for a dissertation or journal manuscript. The analysis method determines the specific evidence and reporting conventions.

The next examples are also illustrative. A real qualitative report must use authentic, consented and appropriately anonymised data extracts from the study.

In an actual paper, ensure that the theme reflects the analysis process and that quotations are not selected merely because they sound persuasive. The data extract should be relevant to the analytic claim.

Journal articles require a concise results section that meets the journal’s reporting expectations. Read recent articles of the same type in the target journal to understand the level of detail, heading conventions, table style and terminology. Do not copy another article’s wording or assume its design is identical to yours.

Prioritise the primary outcome and the findings that directly answer the research question. Use supplementary files for permitted additional detail, but keep essential evidence in the main manuscript. Confirm that the abstract, results, tables, figures and discussion report consistent values and conclusions.

Before submission, check the appropriate reporting guideline, journal instructions, figure resolution, table formatting, anonymisation requirements and word limit. Ensure that all abbreviations are defined and that tables and figures are cited in numerical order.

Peer reviewers may ask for clarification about denominators, missing data, model specification, exclusions or unexpected findings. A transparent results section anticipates these questions by reporting enough information for readers to assess the evidence.

• The section follows the required order and heading structure.
• Every major research question, objective or hypothesis has a corresponding result.
• Primary, secondary and exploratory findings are clearly distinguished.
• All reported numbers match the final analysis output.
• Sample sizes and denominators are stated wherever needed.
• Units, scales, group labels and abbreviations are consistent.
• Effect estimates and suitable measures of uncertainty are reported where appropriate.
• Non-significant, null, unexpected and adverse findings are not selectively omitted.
• Tables and figures are numbered, titled, labelled and cited in order.
• Every table and figure agrees with the narrative and supplementary files.
• Qualitative themes are supported by authentic, anonymised evidence extracts.
• Quotations retain their original meaning and follow the transcription convention.
• Results are not overstated as causal when the design does not support causation.
• Interpretation and implications are reserved for the discussion where appropriate.
• Reporting guidelines and institutional or journal instructions have been checked.
• Confidential and identifying information has been removed.
• The abstract accurately reflects the final results.

Read the section once as a researcher checking accuracy and once as a reader who has not seen the dataset. The first pass catches factual and analytical errors. The second reveals unclear labels, unexplained abbreviations, missing context and gaps in the narrative.