The SPSS output for an independent-samples t-test contains one table with two rows, and picking the wrong one is among the most common errors we see in thesis drafts. If Levene's test came back significant (Sig. below .05), the answer is short: read the bottom row, "Equal variances not assumed." This article explains why, what those decimal degrees of freedom mean, and how to write the result up in APA format.
What Levene's Test Is Actually Checking
The classic Student's t-test assumes the two groups have roughly equal variances, a condition called homogeneity of variance. Levene's test checks exactly that. Its null hypothesis says the variances are equal, so a significant result (p < .05) means they are not. In a comparison of exam scores between 34 full-time and 41 part-time students, a Levene's result of F = 5.62, p = .020 tells you the spread of scores differs meaningfully between the two groups.
Here is the part that trips students up. Levene's test says nothing about your research question. It does not compare means, and its p-value is not your t-test result. It is a gatekeeper with one job: telling you which of the two output rows is valid for your data.
The Two Rows, Decoded
SPSS runs both versions of the t-test every time and prints them stacked in one table. The top row, "Equal variances assumed," is the classic Student's t-test. The bottom row, "Equal variances not assumed," is Welch's t-test, which adjusts the standard error and the degrees of freedom to stay accurate when variances differ.
| Levene's test result | Row to read | Test you are reporting |
|---|---|---|
| Sig. ≥ .05 (not significant) | Top: "Equal variances assumed" | Student's t-test |
| Sig. < .05 (significant) | Bottom: "Equal variances not assumed" | Welch's t-test |
Once you know your row, read across it for the values you need: t, df, Sig. (2-tailed), and the mean difference with its confidence interval. Everything to the right of the Levene's columns belongs to the t-test itself; the two Levene's columns on the far left exist only to make the row decision. Our step-by-step t-test walkthrough shows the full table with annotated screenshots.
Why the Degrees of Freedom Have Decimals
In the bottom row of a study with 75 participants you might see df = 47.3 instead of the expected 73. That is not an error. Welch's method recalculates the degrees of freedom with the Welch-Satterthwaite equation, weighting each group's variance against its sample size, and the result is almost never a whole number. Report it exactly as printed, to one decimal place: t(47.3) = 2.15, p = .037. Rounding it to a whole number, or quietly substituting the df from the top row, are both mistakes an attentive examiner will catch.
The Mistake That Reverses Conclusions
Reading the wrong row is not cosmetic. The two rows can disagree. With unequal group sizes and unequal variances, the top row might show p = .043 while the bottom row shows p = .081, and a thesis built on the wrong row has just reported a significant effect that does not hold. The reverse happens too. This single row choice is one of the errors covered in our review of statistical mistakes that get theses sent back, and it accounts for a surprising share of supervisor corrections in the 300+ projects we have analyzed.
A second version of the same mistake: reporting the Levene's F and p as if they were the main result. If your Results section says F = 5.62, p = .020 where a t-value should be, the analysis reported the doorman instead of the party.
Should You Just Use Welch's Test Every Time?
Some methodologists argue yes. Delacre and colleagues made the case in 2017 that Welch's t-test should be the default because it loses almost no power when variances are equal and stays accurate when they are not, and R made that switch years ago. The two-step ritual (check Levene's, then choose a row) survives mostly because SPSS presents it that way and most university courses still teach it. For your thesis, either approach is defensible. If your supervisor expects the classic procedure, follow the table above. If you simply report Welch's row every time, note that choice in your methodology chapter and cite the rationale. What matters is consistency across all your comparisons, not the philosophy. And if the variance difference reflects heavily skewed data rather than a genuine spread difference, a Mann-Whitney U test may fit better; our guide to parametric versus nonparametric tests covers that decision.
How to Report It in APA Format
When Levene's test is significant, say so briefly, name the correction, and then report the adjusted result. A complete example: Levene's test indicated unequal variances, F = 5.62, p = .020, so degrees of freedom were adjusted. Part-time students scored higher than full-time students, t(47.3) = 2.15, p = .037, d = 0.51, 95% CI [0.28, 8.43]. Two sentences cover the assumption, the fix, and the finding.
When Levene's test is not significant, no disclosure is needed. Report the top row as a standard t-test: t(73) = 2.15, p = .035, d = 0.50.
If you would rather not manage row choices at all, Academic Stats Agent checks variance homogeneity automatically when it runs an independent-samples t-test, applies the Welch correction when needed, and states in the report which version was used and why. Running the analysis is free.
Key takeaway: A significant Levene's test (p < .05) sends you to the bottom SPSS row, "Equal variances not assumed," which is Welch's t-test. Report its t, decimal df, and p exactly as printed, and mention the adjustment in one sentence. The Levene's F and p themselves are never your main result.