A paired-samples t-test compares two measurements taken from the same people, such as before and after an intervention. SPSS prints three tables for it (four in SPSS 27 and later), and two of them contain a column called Sig. Only one of those is your result. This guide reads each table column by column using a worked example, lists the misreadings to avoid, and gives the APA 7 write-up. The reporting format alone is in the paired t-test APA guide.
What SPSS gives you
Analyze > Compare Means > Paired-Samples T Test produces, in this order:
- Paired Samples Statistics: the mean, N, SD and standard error of each measurement.
- Paired Samples Correlations: how strongly the two measurements correlate.
- Paired Samples Test: the t-test on the differences. This is the result.
- Paired Samples Effect Sizes (SPSS 27 and later): Cohen's d and Hedges' correction.
The worked example
Twenty-five employees completed a perceived-stress scale (0 to 40, higher means more stress) before and after an eight-week workplace stress programme. In the dialog, Stress (pre) was entered as Variable 1 and Stress (post) as Variable 2. All numbers below come from one dataset and agree with each other.
Table 1: Paired Samples Statistics
| Mean | N | Std. Deviation | Std. Error Mean | ||
|---|---|---|---|---|---|
| Pair 1 | Stress (pre) | 25.76 | 25 | 5.286 | 1.057 |
| Stress (post) | 22.64 | 25 | 5.163 | 1.033 |
- Pair 1 labels the comparison. With several pairs in one run, each gets its own block.
- Mean: 25.76 before, 22.64 after.
- N: 25 for both. A paired test only uses people with both scores, so N is always the same on the two rows.
- Std. Deviation: 5.29 and 5.16. These are the SDs you report for each time point.
- Std. Error Mean: SD divided by √N; you rarely report it.
Table 2: Paired Samples Correlations
| N | Correlation | Sig. | ||
|---|---|---|---|---|
| Pair 1 | Stress (pre) & Stress (post) | 25 | .565 | .003 |
This table says how strongly the before and after scores are related: people who were stressed before tended to be stressed after (r = .565). Its Sig. (.003) tests whether that correlation is zero. It is not the result of the t-test. It tells you nothing about whether stress went down; a strong correlation is simply why the paired test is more powerful than an independent one. SPSS 27 and later split this Sig. into One-Sided p and Two-Sided p; neither is your result either. Most write-ups do not report this table.
Table 3: Paired Samples Test
In SPSS the first five columns sit under a "Paired Differences" heading, and the CI columns under "95% Confidence Interval of the Difference".
| Mean | Std. Deviation | Std. Error Mean | 95% CI Lower | 95% CI Upper | t | df | Sig. (2-tailed) | ||
|---|---|---|---|---|---|---|---|---|---|
| Pair 1 | Stress (pre) - Stress (post) | 3.120 | 4.876 | .975 | 1.107 | 5.133 | 3.199 | 24 | .004 |
- Mean = 3.120: the average of each person's pre score minus their post score. The label "Stress (pre) - Stress (post)" says which way round SPSS subtracted, Variable 1 minus Variable 2. Positive here means scores fell.
- Std. Deviation = 4.876: the SD of the difference scores, not of either measurement.
- Std. Error Mean = .975: 4.876 / √25.
- 95% CI Lower / Upper = 1.107 to 5.133: the average drop in stress is plausibly between about 1 and 5 points. The interval excludes 0, consistent with p < .05.
- t = 3.199: the mean difference divided by its standard error (3.120 / .975).
- df = 24: the number of pairs minus 1.
- Sig. (2-tailed) = .004: the two-sided p value of the test. In SPSS 27 and later read Two-Sided p under "Significance".
Table 4: Paired Samples Effect Sizes
| Standardizer | Point Estimate | ||
|---|---|---|---|
| Pair 1 | Cohen's d | 4.876 | .640 |
- Standardizer = 4.876, the SD of the differences, SPSS's default for paired data.
- Point Estimate: Cohen's d = 3.120 / 4.876 = 0.64, a medium-to-large effect by Cohen's conventions.
- SPSS also prints a Hedges' correction row (slightly smaller, adjusted for small samples) and a 95% CI for each estimate; only the Cohen's d point estimate is shown above.
Dividing by the SD of the differences gives a different d from dividing by the SD of the scores, and the two are not interchangeable. State which you used, for example "d computed with the SD of the difference scores". Versions before SPSS 27 print no effect sizes: divide the Mean by the Std. Deviation of the Paired Samples Test table.
Misreadings that cost marks
- Reporting the correlation's Sig. as the result. The .003 in Paired Samples Correlations is about the correlation. The test's p is .004, in the Paired Samples Test table.
- Writing p = .000. It means p < .001; write that.
- Reading the sign the wrong way. The Mean is Variable 1 minus Variable 2. If you entered post first, the same result appears as −3.120 and t = −3.199. Say the direction in words.
- Reporting the SD of the differences as a group SD. The 4.876 belongs to the differences; the SDs of the two time points are in Paired Samples Statistics.
- Using the wrong df. It is pairs minus 1 (24), not 2N − 2.
- Running an independent-samples test on paired data. Same people twice means paired; the independent-samples output guide is for two separate groups.
The APA 7 format template
A paired-samples t-test showed that [dependent variable] was significantly higher/lower at [time 2] (M = X.XX, SD = X.XX) than at [time 1] (M = X.XX, SD = X.XX), t(df) = X.XX, p = .XXX, d = X.XX, mean difference = X.XX, 95% CI [X.XX, X.XX].
The write-up for this example
A paired-samples t-test showed that perceived stress was significantly lower after the programme (M = 22.64, SD = 5.16) than before it (M = 25.76, SD = 5.29), t(24) = 3.20, p = .004, d = 0.64, mean difference = 3.12, 95% CI [1.11, 5.13].
The mean difference and its CI are pre minus post, which is why they are positive while the post mean is lower. If a reader could misread that, add "(pre minus post)".
Let KyroStat do the write-up for you
Formatting a paired test by hand means pulling numbers from three tables and keeping the sign straight. KyroStat runs the paired t-test on your spreadsheet, checks the normality of the differences, and writes the APA 7 sentence, table and plot, with the Python or R code behind every number.
Also useful for the write-up (affiliate links): Proofreading for your thesis or paper (20% off with code PAP20): Paperpal
Frequently asked questions
Which Sig. do I report for a paired t-test? The Sig. (2-tailed), or Two-Sided p in SPSS 27 and later, in the Paired Samples Test table. The Sig. in the Paired Samples Correlations table tests the correlation, not the difference.
Do I need to report the paired samples correlation? Usually not. Some reviewers like it because it shows why a paired design was efficient, but the APA write-up of the test does not need it.
Why is the mean difference negative? SPSS subtracts Variable 2 from Variable 1. A negative mean means the second measurement was higher. Swap the order in the dialog, or simply state the direction in words.
What standardizer should Cohen's d use for a paired test? SPSS's default is the SD of the differences, which is the most common choice. Some fields prefer the average SD of the two measurements, which gives a different d: smaller when the two measurements correlate above .5, as they do here, larger when they correlate below .5. Say which one you used.
What if the differences are not normally distributed? With a small sample and clearly non-normal differences, use the Wilcoxon signed-rank test instead.
Which related guides might I need? Often these: paired t-test APA format, repeated-measures ANOVA output for three or more time points, and confidence intervals.