How to Report a Paired-Samples t-Test in APA 7
Academic Decoder (@academicdecoder)
Created by a PhD-qualified lecturer · Updated
The short answer
A paired-samples t-test (also called a repeated-measures or dependent t-test) compares two scores from the same people, such as before and after an intervention. In APA 7 you report the mean and standard deviation at each timepoint, then t(df), the exact p-value, an effect size (Cohen's d, saying which version), and the mean difference with its 95% confidence interval. Use the template that matches your result.
A paired-samples t-test was conducted to compare outcome between Time 1 and Time 2. The analysis revealed a statistically significant increase / decrease in scores from Time 1 (M = mean, SD = SD) to Time 2 (M = mean, SD = SD), t(df) = t, p = p, dav = Cohen's d. The mean difference (Time 1 − Time 2) was difference, 95% CI [lower limit, upper limit].
A paired-samples t-test was conducted to compare outcome between Time 1 and Time 2. Scores were descriptively higher / lower at Time 2 (M = mean, SD = SD) than at Time 1 (M = mean, SD = SD), but this change was not statistically significant, t(df) = t, p = p, dav = Cohen's d. The mean difference (Time 1 − Time 2) was difference, 95% CI [lower limit, upper limit].
If you report dz instead, write dz in place of dav. The confidence interval goes with the mean difference, in its own sentence, so it can't be mistaken for an interval around d.
This guide is for undergraduate and master's students writing a dissertation results chapter. It covers which numbers to take from SPSS, jamovi or R, a worked example, which Cohen's d to report, the formatting rules, and how to word a non-significant result. If your two sets of scores come from different people, you need the independent-samples t-test guide instead.
What to report, and where to find it
You need six values. Most come straight from your software's output; Cohen's d needs a decision about which version to report (see Which Cohen's d?).
M and SD at each timepoint
- What it tells the reader
- The direction and spread of scores
- SPSS
- Paired Samples Statistics table
- jamovi
- Descriptives (under Additional Statistics)
- R
mean()
t
- What it tells the reader
- The mean difference relative to its standard error
- SPSS
- Paired Samples Test, column t
- jamovi
- Statistic
- R
t.test(..., paired = TRUE), t
df
- What it tells the reader
- Number of pairs minus 1
- SPSS
- Column df
- jamovi
- df
- R
- df
p
- What it tells the reader
- Whether the change is statistically significant
- SPSS
- Sig. (2-tailed), or Two-Sided p in newer versions
- jamovi
- p
- R
- p-value
Cohen's d
- What it tells the reader
- How large the change is
- SPSS
- Paired Samples Effect Sizes table (SPSS 27+), which gives dz
- jamovi
- Effect size (tick Cohen's d), which gives dz
- R
mean(diff) / sd(diff)for dz
Mean difference and its 95% CI
- What it tells the reader
- The size of the raw change and a range of plausible values for it in the population
- SPSS
- Paired Differences: Mean; 95% Confidence Interval of the Difference
- jamovi
- Mean difference; Confidence interval
- R
- mean difference; 95 percent confidence interval
Check the order of subtraction. SPSS labels the pair, for example "Before − After", and the mean difference, t and the confidence interval all follow that order. If your software subtracts the other way round, the signs flip but the result is the same.
Worked example
A researcher measures anxiety scores in 28 students before and after an eight-week study-skills programme. Each student has a score at both timepoints. These are sample data for illustration.
| Timepoint | M | SD | n |
|---|---|---|---|
| Before | 24.80 | 6.10 | 28 |
| After | 21.30 | 5.70 | 28 |
Test result: t(27) = 3.56, p = .001. Mean difference (Before − After) = 3.50, 95% CI [1.48, 5.52]. Standard deviation of the differences = 5.20. Cohen's dz = 0.67 (what SPSS shows); Cohen's dav = 0.59.
A paired-samples t-test was conducted to compare anxiety score between Before and After. The analysis revealed a statistically significant decrease in scores from Before (M = 24.80, SD = 6.10) to After (M = 21.30, SD = 5.70), t(27) = 3.56, p = .001, dav = 0.59. The mean difference (Before − After) was 3.50, 95% CI [1.48, 5.52].
What each part does
- The test and the outcome tell the reader what was compared, and that the same people were measured twice.
- The means and SDs give the direction. A reader can see scores fell from Before to After before reading the statistics.
- The verdict states significance and direction together: a statistically significant decrease.
- t(27) = 3.56, p = .001 is the test itself. With 28 pairs, df = 28 − 1 = 27. It is not 54: a paired test counts pairs, not scores.
- dav = 0.59 says how big the change is: moderate by Cohen's conventions. The subscript tells the reader which version this is; the next section explains why SPSS shows 0.67 instead.
- The mean difference and its 95% CI [1.48, 5.52] give the raw change (3.50 points) and a range of plausible values for the population mean change. The interval does not include zero, which agrees with the significant result. Stating the order (Before − After) tells the reader why the difference is positive.
Which Cohen's d?
For paired data there is more than one Cohen's d, and they give different numbers for the same result. The two you are most likely to meet are:
| Version | Calculated as | In the example | Where you see it |
|---|---|---|---|
| dz | Mean difference ÷ SD of the differences | 3.50 ÷ 5.20 = 0.67 | SPSS 27+ and jamovi (labelled Cohen's d) |
| dav | Mean difference ÷ the average of the two timepoints' SDs, (SD₁ + SD₂) ÷ 2 | 3.50 ÷ ((6.10 + 5.70) ÷ 2) = 3.50 ÷ 5.90 = 0.59 | The Result Decoder; common in published research |
dz depends on how strongly the two timepoints are correlated, so it can look large even when the change is small compared with how much people normally vary. dav is on the same scale as the standard deviations you report, so it is easier to compare with other studies. Either is acceptable. What matters is that you say which one you report, in the paragraph or in your method section, and that you don't mix versions across your chapter.
APA 7 formatting rules
| Rule | Correct | Incorrect |
|---|---|---|
| Italicise statistical symbols | t, p, M, SD, d | t, p, M, SD, d (not italic) |
| Report the exact p-value, to two or three decimals | p = .001 | p < .05 |
| Very small p-values use "less than" | p < .001 | p = .000 |
| No leading zero for values that cannot exceed 1 | p = .001 | p = 0.001 |
| Keep the leading zero for values that can exceed 1 | dav = 0.59 | dav = .59 |
| Show which Cohen's d you report | dav = 0.59 | d = 0.59 (version not stated) |
| df is the number of pairs minus 1, as a whole number | t(27) = 3.56 | t(54) = 3.56 |
| Use a true minus sign for negative values | 95% CI [−0.83, 3.43] | 95% CI [-0.83, 3.43] |
| Label the CI and give the order of subtraction | The mean difference (Before − After) was 3.50, 95% CI [1.48, 5.52] | CI (1.48 – 5.52) |
Reporting a non-significant result
A non-significant result is reported with the same values. Only the verdict changes. Sample data: Before M = 23.40, SD = 6.20; After M = 22.10, SD = 5.90; 28 pairs.
A paired-samples t-test was conducted to compare anxiety score between Before and After. Scores were descriptively lower at After (M = 22.10, SD = 5.90) than at Before (M = 23.40, SD = 6.20), but this change was not statistically significant, t(27) = 1.25, p = .222, dav = 0.21. The mean difference (Before − After) was 1.30, 95% CI [−0.83, 3.43].
- Say the data did not provide sufficient evidence of a change. Do not say scores "stayed the same" or "did not change"; a non-significant test does not show that.
- Still report d and the CI. Here the CI runs from scores rising by about 0.8 points to falling by about 3.4 points, which shows how uncertain the estimate is.
Common mistakes
- Using an independent-samples t-test on paired data. If the same people were measured twice, the scores are linked and the paired test is the right one.
- Getting df wrong. df is the number of pairs minus 1 (27 here), not the number of scores minus 2.
- Reading the direction from the sign of t. The sign depends on which timepoint was subtracted from which. Use the means for the direction.
- Not saying which d you report. dz and dav differ, and a reader cannot tell which one you used unless you say.
- Writing p = .000. Report p < .001 instead.
- Reporting the paired correlation as the result. The Paired Samples Correlations table is not the test of whether scores changed.
- Claiming the programme caused the change. Without a comparison group, scores can change for other reasons, such as time passing, practice with the questionnaire, or unusually high scores at the start drifting back towards average. Write "scores decreased", not "the programme reduced anxiety".
- Calling a non-significant result "no change". Say there was insufficient evidence of a change.
What this guide doesn't include
This guide shows you the pattern using sample data. It can't work with your own results, so it doesn't give you:
- A paragraph written for your data. You fill in the templates yourself and check every value by hand.
- A check of your numbers. It can't tell you whether your t, df, p and means agree with each other, or warn you when something looks wrong.
- Cohen's d and the confidence interval if your software doesn't show them.
- A plain-English explanation and a chart of your own result.
- Answers to examiner questions about what your result does and doesn't show.
- The other 21 tests, from ANOVA to regression.
Want this for your own result?
For personalised output, use the Result Decoder. Enter the values from your SPSS, jamovi or R output and get your APA 7 paragraph, a check of your values, a plain-English explanation, charts and help defending your result.
Go to the Result DecoderBe ready to defend it
Supervisors and examiners rarely ask you to recite the numbers. They ask what the numbers allow you to claim. Three questions come up often for a paired t-test. Could you answer each one in a sentence, about your own result?
- "Why did you use a paired test rather than an independent-samples test?"
- "Does this show your programme caused the change?"
- "Which Cohen's d did you report, and why is it different from the one in your SPSS output?"
Practise with your own result
The Result Decoder's Defend this result panel asks questions like these about your own values, shows what your result does and doesn't support, and gives a suggested answer to compare with yours.
Try it with your resultFrequently asked questions
Do I need to report the Paired Samples Correlations table?
It isn't required. The correlation shows how strongly scores at the two timepoints go together; it is not the test of whether the means differ. Some supervisors like it mentioned in a sentence, so check what yours expects.
Which assumption does a paired t-test need?
The differences between each case's two scores should be roughly normally distributed. It is the difference scores that matter, not the scores at each timepoint. With a small sample and clearly non-normal differences, the Wilcoxon signed-rank test is the usual alternative.
What if I have three or more timepoints?
A paired t-test compares two measurements only. For three or more, use a repeated-measures ANOVA rather than running several paired t-tests.
What if some participants are missing a score?
The paired test only uses cases with a score at both timepoints. Report the number of complete pairs, which is df + 1.
Should I report a one-tailed p-value?
Only if you stated a directional hypothesis before collecting data. If you do, say "one-tailed" in the results paragraph.
More statistics guides
Browse all Statistics GuidesDecode your own result
Enter the values from your own SPSS, jamovi or R output to get your APA 7 paragraph, a plain-English explanation and charts, and to practise defending your result. Covers 22 tests for a one-time £15.99.