Education statistics guide
How to Report a t Test in APA 7
A clear APA 7 t-test report names the test, descriptive statistics, t, df, p, and effect size.
Updated:
Problem
Writing only "the groups differed significantly" leaves out the design, degrees of freedom, p value, effect size, and direction.
Who should use this
Use this for papers, assignments, research reports, and SPSS result write-ups.
Formula and concept
Common APA fragment: t(df) = value, p = value, d = value.
Independent-samples t tests compare independent groups. Paired-samples t tests compare repeated or matched observations. Confirm the design before writing the result. A t test is built for exactly two groups or two paired conditions — once you have three or more groups, running several pairwise t tests instead inflates the chance of a false positive, so switch to a one-way ANOVA (see the ANOVA APA 7 guide) rather than stacking t tests. If sample sizes are small and the data are clearly skewed rather than roughly normal, a nonparametric alternative such as the Mann-Whitney U test is usually a safer choice than forcing a t test.
Independent-samples t tests also assume the two groups have roughly equal variances. When Levene’s test (see the SPSS Levene test guide) signals unequal variances, report the Welch-corrected row instead, which typically has a non-integer degrees of freedom rather than a whole number. For p values, use p = .032 style formatting, and write p < .001 instead of p = .000 — follow your course or journal’s exact formatting rules either way.
This page helps with calculation and reporting, but users should still confirm their research design, assumptions, and statistical interpretation.
Step by step
- Confirm whether the test is independent, paired, or one-sample.
- Prepare group means, standard deviations, and sample sizes when needed.
- Copy t, df, and p from the correct output row.
- Add an effect size such as Cohen’s d and state the direction.
Worked example
Independent example: The experimental group (M = 82.40, SD = 7.10) scored higher than the control group (M = 76.30, SD = 8.20), t(58) = 3.07, p = .003, d = 0.79. Paired example: Posttest scores (M = 85.20, SD = 6.40) were higher than pretest scores (M = 78.10, SD = 7.50), t(29) = 5.12, p < .001, d = 0.93. As a check on the independent example: with 30 participants in each group, the pooled degrees of freedom is (30 + 30) − 2 = 58, matching the t(58) reported above. If your own calculation instead lands on a decimal df, you have likely copied the Welch-corrected row rather than the equal-variances row, so confirm which row your software actually reported before writing the final sentence.
Common mistakes
- Using Levene p as the t-test p value.
- Omitting df, or copying the wrong variance row — a quick tell is a non-integer df, since that only comes from the Welch-corrected row, so a df that is not a whole number means you should already be reading the second row, not the first.
- Writing p = .000 instead of p < .001.
- Not saying which group or time point was higher.
Editorial and reproducibility review:
Turn the guide into an auditable decision
This section reworks the guide’s own independent-samples example and uses the two numeric tells it flags — a pooled df that must match group sizes, and a Cohen’s d that looks implausibly large — as self-checks.
Reproduce the independent example and verify df = 58 against the actual group sizes
The example reports t(58) = 3.07, p = .003, d = 0.79 for two groups of 30 each (M=82.40 vs M=76.30). Verify the degrees of freedom independently: pooled df = (30 + 30) − 2 = 58, matching the reported t(58). If your own recomputation instead produces a decimal df, such as 54.32, that is not a rounding artifact — it means you copied the Welch-corrected row rather than the equal-variances row, so go back and confirm which row your software actually reported before writing the sentence.
Rewrite the same numbers as a paired-samples sentence to see why the design word matters
The guide’s paired example is t(29) = 5.12, p < .001, d = 0.93 for pretest/posttest scores from the same 30 students. Compare the two df values side by side: the independent example with 30 per group gives df = 58, while the paired example with 30 students gives df = 29 (n − 1). Same sample size, different design, different formula for df — confirm which design you actually ran before copying either df pattern into your own report.
Use the plausibility checks the guide flags for d and for row selection
Self-test with the reported d = 0.79: if your own calculation instead produces something like d = 2.3 for a simple two-group comparison, that is implausibly large, and the likely cause is dividing by one group’s SD instead of the pooled SD. Similarly, a non-integer df by itself (not 58, but something like 54.3) tells you that you are already reading the Welch-corrected row, so a df that is not a whole number means you should not also be copying the equal-variances row’s t and p values.
Completion criteria
- Reproduced the independent example and verified pooled df = (30+30)−2 = 58 matches the group sizes.
- Rewrote the same scenario as a paired-samples sentence and can explain why df = 29 (n−1) differs from the independent df = 58 for the same sample size.
- Would flag a Cohen’s d above roughly 2 for a simple two-group comparison as a signal to recheck whether the pooled SD was used.
- Can state that a non-integer df on its own indicates the Welch-corrected row, not the equal-variances row.
Recommended tools
Related guides
FAQ
- Do I need Cohen’s d?
- Many courses and journals require an effect size, and it helps explain the size of the difference. If the value looks implausibly large, such as d > 2 for a simple two-group comparison, recheck whether you divided by the pooled standard deviation rather than by a single group’s SD.
- Can I copy p = .000 from SPSS?
- No. Write p < .001.
- Can df be a decimal?
- Yes, especially with Welch t tests.
- Where should descriptive statistics go?
- Put them in the same sentence or immediately before the test result.
Next step
Check the SPSS row first, then use the APA 7 report generator for an editable sentence.