A one-way ANOVA in SPSS can print six or seven tables, and the one that answers "which groups differ?" is three tables down from the one that answers "do any groups differ?". This guide reads every table row by row with a worked example (three groups, Tukey HSD post hoc tests), flags the misreadings that cost marks, and ends with the APA 7 write-up. For the reporting format on its own, see the one-way ANOVA APA guide.
What SPSS gives you
Analyze > Compare Means > One-Way ANOVA, with Descriptive and Homogeneity of variance test ticked under Options, Tukey ticked under Post Hoc, and Estimate effect size ticked (SPSS 27 and later), produces:
- Descriptives: N, mean, SD, standard error, 95% CI, minimum and maximum per group.
- Tests of Homogeneity of Variances: Levene's test.
- ANOVA: the omnibus F test. This is the main result.
- ANOVA Effect Sizes (SPSS 27 and later): eta squared and related measures.
- Multiple Comparisons: every pairwise Tukey comparison.
- Homogeneous Subsets: the same Tukey results, grouped.
The order can differ slightly between versions; the table titles are what to look for.
The worked example
Sixty students were randomly assigned to one of three ten-minute breaks before an attention test (scored 0 to 100): no break, a short walk, or scrolling on their phone. Twenty students per group. All tables below come from one dataset.
Table 1: Descriptives
| N | Mean | Std. Deviation | Std. Error | 95% CI Lower Bound | 95% CI Upper Bound | Minimum | Maximum | |
|---|---|---|---|---|---|---|---|---|
| No break | 20 | 49.50 | 5.826 | 1.303 | 46.773 | 52.227 | 37 | 59 |
| Walk | 20 | 60.30 | 7.234 | 1.618 | 56.915 | 63.685 | 46 | 73 |
| Phone | 20 | 51.80 | 6.895 | 1.542 | 48.573 | 55.027 | 36 | 63 |
| Total | 60 | 53.87 | 8.064 | 1.041 | 51.783 | 55.950 | 36 | 73 |
- N, Mean, Std. Deviation: what goes into your descriptives table or text. The walk group scored highest (60.30), the no-break group lowest (49.50).
- Std. Error: SD / √N, not usually reported.
- 95% Confidence Interval for Mean (Lower Bound, Upper Bound): the CI around each group's own mean. It is not a test of differences between groups.
- Minimum, Maximum: a quick check for impossible values or data-entry errors.
- Total: all 60 students together.
Table 2: Tests of Homogeneity of Variances
| Levene Statistic | df1 | df2 | Sig. | |
|---|---|---|---|---|
| Based on Mean | .207 | 2 | 57 | .814 |
Levene's test checks whether the groups have similar variances, an assumption of the standard F test and of Tukey HSD. Sig. = .814 is well above .05, so the assumption holds. Recent SPSS versions print four rows (Based on Mean, Based on Median, Based on Median and with adjusted df, Based on trimmed mean); the "Based on Mean" row is the classic Levene's test shown here, and the median-based row (the Brown-Forsythe version) is more robust when the data are skewed. This Sig. is not your ANOVA result.
Table 3: ANOVA
| Sum of Squares | df | Mean Square | F | Sig. | |
|---|---|---|---|---|---|
| Between Groups | 1294.533 | 2 | 647.267 | 14.512 | .000 |
| Within Groups | 2542.400 | 57 | 44.604 | ||
| Total | 3836.933 | 59 |
- Between Groups row: the variation between the group means. Its df = number of groups − 1 = 2. F = 14.512 and Sig. = .000 belong to this row. This is the result you report.
- Within Groups row: the variation of scores inside the groups (the error term). Its df = N − number of groups = 57 is the second df you report.
- Mean Square = Sum of Squares / df. F is the Between Groups mean square divided by the Within Groups mean square: 647.267 / 44.604 = 14.51.
- Total: Between + Within. Not reported.
Sig. = .000 means p < .001. The exact value here is about .000008; SPSS rounds anything below .0005 to .000.
Table 4: ANOVA Effect Sizes
| Point Estimate | |
|---|---|
| Eta-squared | .337 |
| Epsilon-squared | .314 |
| Omega-squared Fixed-effect | .311 |
- Eta-squared = Between Groups SS / Total SS = 1294.533 / 3836.933 = .337. Break type accounts for about 34% of the variance in attention scores in this sample.
- Epsilon-squared and Omega-squared Fixed-effect correct eta squared for its upward bias in small samples, so they are a little smaller (.314, .311). Omega squared is the better estimate of the population effect.
- SPSS also prints an Omega-squared Random-effect row (for random factors, rarely needed) and a 95% CI for each estimate; only the point estimates are shown above.
Before SPSS 27 there is no effect-size table; compute eta squared from the ANOVA table as Between Groups SS divided by Total SS.
Table 5: Multiple Comparisons (Tukey HSD)
| (I) Break | (J) Break | Mean Difference (I-J) | Std. Error | Sig. | 95% CI Lower Bound | 95% CI Upper Bound |
|---|---|---|---|---|---|---|
| No break | Walk | -10.800* | 2.112 | .000 | -15.882 | -5.718 |
| Phone | -2.300 | 2.112 | .525 | -7.382 | 2.782 | |
| Walk | No break | 10.800* | 2.112 | .000 | 5.718 | 15.882 |
| Phone | 8.500* | 2.112 | .000 | 3.418 | 13.582 | |
| Phone | No break | 2.300 | 2.112 | .525 | -2.782 | 7.382 |
| Walk | -8.500* | 2.112 | .000 | -13.582 | -3.418 |
- (I) Break and (J) Break: every group is compared with every other group, in both directions. The six rows hold three comparisons. No break vs Walk and Walk vs No break are the same test, with the sign of the difference flipped.
- Mean Difference (I-J): the (I) mean minus the (J) mean. An asterisk marks differences significant at .05 (the table's footnote says "The mean difference is significant at the 0.05 level").
- Std. Error: the same for every pair here because the groups are equal in size.
- Sig.: the Tukey-adjusted p value for that pair. Already corrected for multiple comparisons; do not adjust it again.
- 95% Confidence Interval: the Tukey-adjusted CI of the difference. It excludes 0 exactly when Sig. is below .05.
Read once per pair: Walk scored 10.80 points higher than No break (p < .001) and 8.50 higher than Phone (p < .001); No break and Phone did not differ (2.30 points, p = .525).
Table 6: Homogeneous Subsets
| Break | N | Subset 1 | Subset 2 |
|---|---|---|---|
| No break | 20 | 49.50 | |
| Phone | 20 | 51.80 | |
| Walk | 20 | 60.30 | |
| Sig. | .525 | 1.000 |
This table sorts the groups by mean and puts groups that do not differ significantly in the same column. No break and Phone share Subset 1; Walk sits alone in Subset 2. The Sig. at the bottom of each column is the p value for the groups within that subset, so it should be above .05 (a subset holding one group always shows 1.000). SPSS adds footnotes saying that means for groups in homogeneous subsets are displayed and that the harmonic mean sample size is used (20 here). The table repeats what the Multiple Comparisons table says and is not usually reported.
Misreadings that cost marks
- Reporting Levene's Sig. as the ANOVA result. .814 is about variances; the ANOVA's p is in the ANOVA table.
- Writing p = .000. Write p < .001. That applies to the Tukey Sig. values too.
- Reporting only one df. An F needs both: Between Groups df and Within Groups df, F(2, 57).
- Counting six comparisons. The Multiple Comparisons table lists each pair twice. Three groups give three comparisons.
- Stopping at the omnibus test. A significant F says the means are not all equal; Tukey says which pairs differ.
- Using Tukey when variances are unequal. If Levene's test is significant, report Welch's ANOVA (tick Welch under Options) and Games-Howell post hoc tests instead.
- Reading the group CIs in Descriptives as tests. Overlapping 95% CIs of two means do not show that the means are not significantly different; use the Tukey table.
The APA 7 format template
A one-way ANOVA showed a significant effect of [independent variable] on [dependent variable], F(df1, df2) = X.XX, p = .XXX, η² = .XX. Tukey HSD post hoc comparisons showed that [group A] (M = X.XX, SD = X.XX) scored significantly higher than [group B] (M = X.XX, SD = X.XX), p = .XXX.
The write-up for this example
A one-way ANOVA showed a significant effect of break type on attention scores, F(2, 57) = 14.51, p < .001, η² = .34. Tukey HSD post hoc comparisons showed that students who took a walk (M = 60.30, SD = 7.23) scored significantly higher than students who took no break (M = 49.50, SD = 5.83), p < .001, and than students who used their phone (M = 51.80, SD = 6.89), p < .001. The no-break and phone groups did not differ significantly, p = .525.
You can add the Tukey mean differences and their CIs, for example "mean difference = 10.80, 95% CI [5.72, 15.88]", when your field expects them. The post hoc tests guide shows the table version.
Let KyroStat do the write-up for you
Already have the output? Copy the ANOVA table from the SPSS Viewer and paste it into the box on the one-way ANOVA reporting guide: it returns the APA sentence with eta squared and checks that F, the dfs and p agree. Or upload the spreadsheet to KyroStat, which runs the ANOVA, the assumption checks and the post hoc tests, and writes the sentences and tables with the code behind them.
Also useful for the write-up (affiliate links): Proofreading for your thesis or paper (20% off with code PAP20): Paperpal
Frequently asked questions
Which row of the ANOVA table do I report? The Between Groups row gives F, Sig. and the first df; take the second df from the Within Groups row. Report F(Between df, Within df).
Why does each comparison appear twice in the Multiple Comparisons table? SPSS compares every group with every other group from both sides, so A vs B and B vs A both appear with opposite signs. Report each pair once.
Do I report the Homogeneous Subsets table? Usually not. It groups the same Tukey results by subset; the Multiple Comparisons table already has the p values and CIs you need.
What if the ANOVA is significant but no Tukey comparison is? It happens, especially with p values near .05: Tukey controls for several comparisons and the omnibus test does not. Report both results honestly; do not hunt for another post hoc test until one is significant.
What if Levene's test is significant? Use the Welch row of the Robust Tests of Equality of Means table (tick Welch under Options) and Games-Howell post hoc tests.
Which related guides might I need? Often these: eta squared, repeated-measures ANOVA output for the same people measured several times, and the Kruskal-Wallis test as the non-parametric alternative.