SPSS guide

Mann-Whitney U or t-test: which should I use?

Short answer

Use an independent-samples t-test when your outcome is a Scale variable that is roughly normal in both groups. Use the Mann-Whitney U test when the outcome is ordinal (such as a rating) or clearly non-normal, especially with small groups. Both compare two separate groups of people.

The decision in three questions

  • Are the two groups made up of different people? If not, you need the paired t-test or Wilcoxon instead.
  • Is the outcome Scale (scores with equal intervals)? If it is ordinal, use Mann-Whitney U.
  • Is it roughly normal in each group? Check with Explore. If it is clearly skewed or has outliers, particularly with fewer than about 30 per group, use Mann-Whitney U.

What each test compares

The t-test compares means, so it uses the actual size of every score. Mann-Whitney ranks all the scores and compares the ranks, so an extreme value counts no more than any other high score. That makes it robust to outliers but slightly less powerful when the t-test’s assumptions are met.

How to report each one

t-test: report M and SD for each group, then t(df), p and Cohen’s d. Mann-Whitney: report the median for each group, then U, z, p and r.

Common mistakes

  • Switching to Mann-Whitney only because the t-test wasn’t significant.
  • Reporting means and SDs alongside a Mann-Whitney test.
  • Running either test on paired data.

Step-by-step tutorials

Still not sure which test fits?

Answer a few questions and the free tool tells you the test, the SPSS menu path and an APA example.

More SPSS questions

All SPSS tutorials →