How to write results section of a research paper

The results section reports what your study found, using the methods you described before it. The writing guide of the University of Southern California library puts it simply: the results section should "arrange and describe the findings in a logical sequence without bias or interpretation."

results sectionnoun

The part of a research paper or thesis that reports the findings of the study, in a logical order, without explaining why they happened or what they mean.

The most common confusion is between the results and the discussion. Grad Coach sums up the line between them in one sentence: "the results chapter describes the data, while the discussion chapter explains what they mean."

Many journals, and many theses in science and engineering, combine the two into one "Results and discussion" section or chapter. Whether you may combine them depends on your journal or your university, so check the guide for authors or your thesis format first. Even in a combined chapter, keep the report of each finding clearly apart from your explanation of it. For where the results sit among the other parts of a paper, see the structure of a research paper.

Guides from USC, Grad Coach and Researcher.Life suggest much the same order. Before you start, write your research questions or hypotheses on one sheet and keep it beside you. Every paragraph of the results should answer one of them.

1. Remind the reader of the question.One or two sentences restating the aim or the research questions. After the methods, this tells the reader what to look for.
2. Describe what you ended up with.How many people answered, how many samples were analysed, and anything lost or left out along the way. Joel wrote that 30 samples were collected and all 30 were analysed.
3. Report any checks on the data.If your method calls for them, report checks such as the reliability of a questionnaire scale. Grad Coach notes that Cronbach's alpha is the most common test for this.
4. Answer each question in turn.Take the research questions or hypotheses in the same order as your introduction, under clear subheadings, and give each one a plain answer: supported or not supported.
5. Close with a short summary.A few lines that pull the main findings together and lead the reader into the discussion.

Report only what answers your questions; the USC guide warns against findings that do not address the research problem. Do not repeat your methods either. That is the job of the chapter before, explained in how to write a research methodology.

Start with descriptive statistics. Grad Coach lists means, medians and standard deviations for numbers such as scores or counts, and frequencies or percentages for categories such as gender or beach type. Then report the inferential tests that answer your questions. "The most important thing," Grad Coach writes, "is to use the tests that answer your research questions, and to report them fully."

Reporting a test fully means giving more than the word "significant". Put these together in one sentence:

• the test statistic and its degrees of freedom, for example t(28) = 2.42;
• the exact p value, for example p = .022, rather than only "p < .05";
• an effect size, which tells the reader how big the difference is;
• a confidence interval, where your field expects one.

Very small p values are written as p < .001, never as p = .000, which some software prints. House styles differ: APA style drops the zero before the decimal point (.022), while the New England Journal of Medicine keeps it (0.02), as GraphPad's guide to reporting P values shows. Follow your journal or university, and be consistent.

Notice what the rewrite leaves out: the word "proves" and any idea of why. Those questions belong to the discussion. If your study tested a hypothesis, say plainly whether the result supported it; how to write a hypothesis shows how a hypothesis and its result should line up.

Use a table when the reader needs exact values, a figure when the shape of the data matters more than the exact numbers, and plain text when there are only one or two numbers to give. Whatever you choose, show each set of data only once. Joel's first draft had the same six averages in a table, in a bar chart and in the text. He kept the table, dropped the chart and wrote one sentence about the pattern.

• Number every table and figure in the order they appear, and give each a clear title or caption.
• Mention each one in the text before it appears, and say in words what the reader should notice.
• Do not repeat every value from a table in your sentences. Point out the main finding and let the table hold the details.
• Build your own tables. Grad Coach advises against pasting raw software output, because "raw output is cluttered."
• Move long tables, raw data and interview transcripts to an appendix, and refer to it in the text.

Leaving out the sixth beach was the most serious mistake in Joel's first draft, because a reader who later finds missing data will doubt everything else. The USC guide lists ignoring negative results among the things to avoid: report them, and explain in the discussion why they might have happened.

Grad Coach says the same of hypotheses: "one hypothesis at a time, each with a verdict", supported or not.

Joel put the river mouth beach back. In the rewrite, one plain sentence reports it: the quiet beach near the river mouth had the highest mean count of all six beaches. Its possible explanation, plastic carried down by the river, waits for the discussion. Leaving data out to make a result look cleaner misrepresents what the study found.

In a qualitative study, the results are usually called findings, and they are organised by themes instead of tests. Joel also interviewed eight people who run beach shacks about how the sand is cleaned. Guides on writing up qualitative research agree on a few points:

1. Use your main themes as subheadings. A label taken from the participants' own words often works well; one of Joel's became "the river brings it".
2. Open each theme with your own statement of the pattern, then give the quotes that support it.
3. Give each quote enough context to be understood. Use "..." where words are cut and [square brackets] for words you add.
4. Spread the quotes across participants, and say where views differed. Two talkative people should not stand in for eight.
5. Use codes or pseudonyms, and leave out details that could identify someone, such as the exact name and location of a shack.

Report what participants said and how widely a view was shared, and keep your explanation for the discussion. For how the two kinds of study differ, see qualitative vs quantitative research.

Here is a short part of Joel's rewritten results, under the subheading for his first research question:

All 30 sand samples (five from each of six beaches) were analysed. Microplastic pieces were found in every sample. Busy beaches had more pieces per kilogram of dry sand (M = 41, SD = 12) than quiet beaches (M = 29, SD = 15), t(28) = 2.42, p = .022, d = 0.88. The hypothesis was therefore supported. However, counts varied widely within the quiet group: the beach nearest the river mouth had the highest mean count of all six beaches (Table 2).

It names the sample, gives the numbers once, reports the test fully, states the verdict on the hypothesis, includes the result that did not fit and points to the table. It does not say why. That is the job of the discussion.

These are the mistakes Joel's first draft made, and the ones examiners and reviewers notice most:

• explaining or judging the findings instead of reporting them;
• pasting software output instead of building clean tables;
• showing the same numbers in a table, a chart and the text;
• leaving out results that did not fit the expected pattern;
• writing "significant" with no test, no exact p value and no effect size;
• adding results that do not answer any research question.

Joel's new results chapter was six pages shorter than the first. At his viva, the external examiner turned to Table 2 and asked about the river mouth beach. Joel was ready: the number was in his results, and his discussion had a paragraph on why it might be so.