From Output to Results Section
Statistical software gives you pages of output. The Results section turns that output into a few plain, precise sentences, one per analysis, that a reader can check and reproduce. Every results sentence follows the same formula, so once you know it, writing them becomes mechanical. The harder part is knowing what to leave out: interpretation, spin, and the tempting shortcuts that break APA style.
The results-sentence formula
Almost every sentence in a Results section is built from the same five moves, in this order:
- What was tested, in the variables' own words: "the nap group recalled more word pairs than the no-nap group". Name the comparison and, where it applies, the direction.
- The descriptives that anchor it: group means with SDs in parentheses (M = 28.4, SD = 5.1), so the reader sees the raw size of the difference.
- The test statistic with its degrees of freedom and exact p: t(58) = 3.17, p = .002. The df encode your n and design, and a grader reads them first (see Reporting Statistics in APA Style).
- The effect size (d = 0.82, η² = .11, r = .34), so "significant" comes with "how big".
- The confidence interval, wherever one is available: 95% CI of the difference [1.59, 7.01]. APA 7 asks for it, and it is useful because it gives the whole range of effects your data are compatible with. The effect size says how big the difference was in this sample; the interval says how precisely you have estimated it.
Put them together and a whole analysis fits in one sentence: "The nap group recalled more word pairs (M = 28.4, SD = 5.1) than the no-nap group (M = 24.1, SD = 5.4), t(58) = 3.17, p = .002, d = 0.82, 95% CI of the difference [1.59, 7.01]." Notice what's not there: no "which proves that sleep helps memory". That's the Discussion's job.
Report the analysis you planned
The Results section reports the test you said you would run, ideally the one in your preregistration or analysis plan. If you also ran extra, unplanned analyses, that's fine, but label them exploratory so the reader can weight them accordingly. Silently swapping the planned test for whichever one crossed p < .05, or reporting only the comparisons that "worked", leads to results that don't replicate. Confirmatory and exploratory analyses are both legitimate. The only rule is that you say which is which.
Results report; the Discussion interprets
The most common Results-section error is including interpretation. The Results section says what happened; the Discussion says what it means. "Recall was higher in the nap group, t(58) = 3.17, p = .002" is a result. "…which shows that sleep consolidates memory" is a claim about the world. It belongs in the Discussion, where you can hedge it, relate it to theory, and admit its limits.
Two more house rules that trip students up:
- Past tense. You are reporting things that already happened ("scores differed", "the model explained 12% of the variance"), not "scores differ".
- Assumption checks, briefly. Mention that you checked (normality, homogeneity of variance, etc.) and what you did if they failed, in a clause or two, not a page. "Levene's test indicated equal variances (p = .41)" is plenty.
Build the results sentence
Here are three analyses as your software would report them. For each, click the fragments that belong in a clean APA results sentence, and leave out the traps (wrong df, p = .000, a stray leading zero, a sneaky interpretation). The sentence assembles as you go.
🧩 Assemble the results sentence
Pick the correct pieces from the output; avoid the traps.
Click every fragment that belongs in the sentence:
One analysis, one sentence, five parts. What you tested → the descriptives → the statistic with df and exact p → the effect size → the interval, where the output gives you one. If a piece you're about to add is a claim about meaning, or a number that doesn't come straight off the output, it doesn't belong in Results.
Why it matters: graders, reviewers, and anyone trying to replicate your study check the Results section most closely, because it can be checked. The df must match your n, the p must match the statistic, and every planned analysis must be there. A tidy, complete Results section with no interpretation suggests that the rest of the paper can be trusted too. Once you know the formula well, this is the easiest section to write. For annotated output of the five classic tests, with the reportable cells highlighted, see the guide From SPSS Output to APA Results.
Problem 70 of the practice problems gives you a trial where twenty-four people never came back, unevenly across the two arms, and asks what the Results section is obliged to say about them.
Common questions
How do I turn my statistical output into a results sentence?
Use the same five parts every time, and each analysis becomes one sentence. Begin with what you tested, naming the variables and which way the difference went. Add each group's descriptives, the mean followed by the SD in parentheses. Then report the test: the statistic, its degrees of freedom and the exact p. End with the effect size and, if your output gives one, the confidence interval. Put together, they read: "The nap group recalled more word pairs (M = 28.4, SD = 5.1) than the no-nap group (M = 24.1, SD = 5.4), t(58) = 3.17, p = .002, d = 0.82, 95% CI of the difference [1.59, 7.01]." With the formula in hand, the Results section is the easiest one to write. See Reporting Statistics in APA Style for the typography.
Should I report exact p-values or just p < .05?
Report exact p-values wherever you can: p = .023 tells the reader far more than 'p < .05', and modern APA style asks for it. Round to two or three decimals, and use the < sign only at the floor: anything below .001 is written p < .001. Never write p = .000, because a p-value is never exactly zero; that output just means 'too small to display'. The one time a threshold form is acceptable is a table footnote convention (e.g. * p < .05), but in the running text, give the number.
How do I report participants who dropped out or were excluded?
With numbers and a stated rule, not with adjectives. Say how many people started, how many finished, and how many you removed and on what criterion. Then give the n you actually analyzed, so a reader can follow every participant from recruitment to the final test. The criterion matters as much as the count: "six participants scoring below chance were removed" can be checked and reproduced, while "outliers were removed" cannot. If the loss falls unevenly across conditions, say so explicitly, because that pattern is itself a finding and it can bias the comparison in a way no test repairs. Two habits keep exclusions from being driven by the results. Fix the exclusion rule before you look at the outcome, ideally in your preregistration. And if a judgment call was involved, report the analysis both with and without those cases.