How to report an independent-samples t-test in APA style (7th ed.)
A two-group comparison is the most common test in a thesis, and the most common sentence to get wrong. Here is what the sentence needs, the pattern, a worked example, and the six mistakes reviewers flag.
What the sentence must contain
The two group means and standard deviations; the test statistic with its degrees of freedom; the p-value; an effect size (Cohen’s d) with its confidence interval; and the direction of the difference in words. A sentence that stops at "p < .05" tells the reader that something differed but not by how much, or which way.
The pattern
Group A (M = , SD = ) scored higher than Group B (M = , SD = ), t(df) = x.xx, p = .xxx, d = x.xx, 95% CI [ , ].
Italicise M, SD, t, p and d. Report t and d to two decimals, p to three (or "p < .001"). Drop the leading zero only for values that cannot exceed 1 — p, r — and keep it for d, which can.
A worked example
On a synthetic sample of 60 respondents, F (M = 4.11, SD = 0.66) scored higher than M (M = 3.39, SD = 0.71), t(58) = 4.06, p < .001, d = 1.05, 95% CI [0.36, 1.07]. Levene's test indicated equal variances.
This sentence was produced by the engine that computed the table; you can walk the same sample through in six steps.
Equal variances, and what to do when they are not
Say which t you report. If Levene’s test is not significant, Student’s t with df = n₁ + n₂ − 2. If it is, report Welch’s t and its non-integer degrees of freedom — t(47.3) = 2.14 — and say so. Reporting Levene only when it fails leaves the reader guessing which test produced the number.
Six mistakes reviewers flag
- p = .000 — write p < .001. Software prints zeros; probabilities are never zero.
- No effect size — d (or its CI) is required by APA 7, not optional.
- "Highly significant" — significance has no degrees; report the value.
- Direction missing — "differed significantly" without saying which group was higher.
- Rounded df for Welch’s t — keep the decimal.
- Means without SDs — a mean alone cannot be compared across studies.
Where the numbers come from
In Statory the sentence is generated from the same computation that produced the table, so the two cannot disagree. Every engine is checked line by line against established statistics software and the comparison is published — see the accuracy page.
The same engine runs inside ChatGPT and Claude when you paste a table there.