Written and maintained by CASRAI Editorial Board
Last updated
Most SPSS guides teach you how to run one specific procedure. This one is different: it covers the tables the SPSS Viewer prints for almost any procedure, so you can read a colleague’s output, a dataset you inherited, or your own results from six months ago, without re-learning the mechanics of whichever test produced them. If you can already run the test you need, the individual procedure guides linked throughout this page cover that; this page is about what the numbers on the screen actually mean once you have them.
Where SPSS output actually lives
Every procedure you run writes its results to the Output Viewer (a separate window from the Data Editor), as a scrolling, navigable log: a syntax echo of what you ran, one or more tables, and any charts, in the order you ran them. The whole session can be saved as an .spv file or exported to Word, PDF, HTML, or Excel for a manuscript or report — see What Is SPSS? for how this fits into the wider workflow. The tables themselves are not static images. They are Pivot Tables, and double-clicking one switches it into an editable state where you can reorder rows and columns, hide footnotes, or reveal more decimal places than the default display shows. That last one matters more than it sounds — see the next section.
The rounding trap: why “Sig. = .000” never actually means zero
SPSS displays most numeric output to three decimal places by default. A p-value smaller than .0005 rounds down to .000 on screen — it is not, and cannot be, exactly zero. Double-click the table and increase the decimal places in the cell’s properties (or widen the column) and you will see the real value, typically in scientific notation (e.g. 2.45E-006). This is a display artifact of every SPSS output table with a Sig. column, not a property of any one test.
The correct way to write this up is not “p = .000” — report it as p < .001. See How to Report P Values: APA Rules for the full convention, including how to handle values that round to exactly .050. The same three-decimal rounding applies to test statistics, correlation coefficients, and confidence interval bounds throughout SPSS output — if a number looks suspiciously round, it is worth checking whether that is the true value or just the display.
Case Processing Summary: the table you didn’t ask for
Many procedures — crosstabs, reliability analysis, several nonparametric tests — print a Case Processing Summary table before the results you actually wanted. It reports how many cases were Valid, how many were Excluded (almost always due to missing data on one of the variables involved, under SPSS’s default listwise deletion), and the Total, each as a count and a percentage. Its job is quality control, not analysis: check that the Valid N matches what you expect before trusting anything below it. A Valid N that is much smaller than your dataset’s total row count is the single most common silent error in SPSS output — it means a variable had more missing data than you realized, or a filter is still active from an earlier step (Data > Select Cases).
Footnote superscripts: a, b, c aren’t decoration
Nearly every SPSS table that isn’t a simple descriptives list carries small superscript letters (a, b, c…) next to specific cells or row labels, each tied to a footnote printed directly below the table. These are not optional annotations — they carry information the table’s numbers alone don’t state, and they differ by procedure:
- In a regression coefficients table, a footnote after the table title typically states the dependent variable (e.g. “a. Dependent Variable: score”).
- In an independent-samples t-test, the row label footnotes distinguish “Equal variances assumed” from “Equal variances not assumed” — which row to read depends on Levene’s Test for Equality of Variances, printed in the same table.
- In a one-way ANOVA post hoc table, footnotes identify which correction method (Tukey, Bonferroni, Games-Howell) produced the rows below.
- In a correlation matrix, asterisk footnotes state the significance threshold and whether the test was one- or two-tailed (see below).
The general rule: never read a number out of context of its own table’s footnotes. The same statistic (a mean, a p-value, a correlation) can mean something different depending on which footnote it’s attached to.
Descriptives-style tables: Mean, Std. Deviation, Std. Error
Tables produced by Analyze > Descriptive Statistics and the descriptives block that many other procedures print alongside their main result share a common set of columns: N (valid cases), Mean, Std. Deviation (the sample standard deviation, not the population one), Minimum, Maximum, and sometimes Std. Error of the Mean (the standard deviation divided by the square root of N — a measure of how precisely the sample mean estimates the population mean, not a measure of spread in the data itself). See Descriptive Statistics in SPSS for how to generate these and the difference between the Frequencies, Descriptives, and Explore procedures that each produce a slightly different version of this table.
Test-statistic tables: the columns that repeat across ANOVA, regression, and t-tests
Despite looking different at a glance, the tables for a t-test, a one-way ANOVA, and a regression F-test share a common logical structure, because they’re all testing whether an observed effect is larger than chance variation would produce:
- df (degrees of freedom) — tied to sample size and the number of groups or predictors; needed to interpret the test statistic and required in APA-style reporting.
- The test statistic itself — t, F, or χ², depending on the test — how far the observed data sits from what the null hypothesis predicts, in standardized units.
- Sig. (or Sig. (2-tailed)) — the p-value; see the rounding section above before reporting it directly from the screen.
- An effect-size column, when present — partial eta squared in ANOVA output, Cohen’s d in some t-test add-ons, R² in regression’s separate Model Summary table. Statistical significance and effect size answer different questions (whether an effect exists at all vs. how large it is), and SPSS reports them in different tables or columns — see effect size and Cohen’s d for why both belong in a results section.
The procedure-specific guides on this site (t-test, one-way ANOVA, multiple regression) walk through what each of these means for that specific test; this page is about recognizing the pattern once you already know it’s there.
Correlation tables: the one place SPSS prints asterisks by default
Most SPSS output reports exact Sig. values with no symbolic flagging — you decide significance by comparing the number to your chosen alpha. The Bivariate Correlations table (Analyze > Correlate > Bivariate) is the main exception: with the default “Flag significant correlations” option left checked, SPSS marks each correlation coefficient with a single asterisk (*) if it is significant at the .05 level, or a double asterisk (**) at the .01 level, with the exact rule stated in a footnote below the table (“Correlation is significant at the 0.05 level (2-tailed).”). It is worth knowing this is an exception rather than the general SPSS convention — a habit of expecting asterisks everywhere leads to skimming past an unflagged but genuinely significant result in, say, a regression coefficients table. See Pearson Correlation in SPSS for the full procedure.
Model Summary and Coefficients: regression’s two-table split
Regression output is worth calling out on its own because it’s split across (at minimum) three tables that each answer a different question, and reading only one of them is a common source of an incomplete write-up: the Model Summary table (R, R Square, Adjusted R Square — how much variance the model as a whole explains), an ANOVA table (whether the model as a whole is a statistically significant predictor of the outcome), and a Coefficients table (B, Std. Error, Beta, t, Sig., and confidence intervals for each individual predictor). A model can be significant overall (the ANOVA table) while a specific predictor within it is not (its row in the Coefficients table), and vice versa in a multi-predictor model — both tables need to be read, not just one.
Getting a table into a manuscript
Because SPSS output tables are Pivot Tables, they can be copied directly into Word (right-click > Copy, then paste as a Word table, not as a picture, to keep it editable) or exported in batch via File > Export. Neither approach produces an APA-formatted table automatically — SPSS’s default table borders, labels, and column headers do not match APA style, and journals routinely reject a pasted-in raw SPSS table. Treat the Viewer’s table as the source data to build your manuscript table from, not as the manuscript table itself.
Frequently asked questions
Why does SPSS give two p-values for some tests?
Where a table reports a one-tailed and two-tailed Sig. value (some nonparametric tests, some correlation output), use the two-tailed value unless your hypothesis specifically predicted the direction of the effect in advance — a two-tailed test is the standard default and the more conservative choice.
Why is my Valid N different across two tables from the same analysis?
Each table applies listwise deletion based on the variables involved in that specific piece of output. A descriptives table for one variable can have a higher N than a correlation table built from that same variable paired with a second one that has more missing data.
Can I change how many decimal places SPSS shows?
Yes — double-click into the Pivot Table, select the cell or column, and use Cell Properties (or Format > Cell Properties) to increase the displayed decimals. This changes the display, not the underlying stored precision, which SPSS always keeps regardless of what’s shown.
What does an empty cell or a period (.) in a table mean, instead of a number?
A period in a numeric cell (distinct from a footnote symbol) generally means the value is not applicable to that row or column — for example, a diagonal cell in a correlation matrix comparing a variable to itself, which SPSS leaves blank of a Sig. value rather than printing a meaningless “1.000, p = —”.
For the mechanics of any one specific test — how to run it, which menu path to use, and what its own output table specifically means — see the individual SPSS guides linked throughout this page, or the full list at Research Methods. For writing SPSS analyses up so someone else (including a future version of you) can re-run them, see SPSS Syntax: Writing Reproducible Analysis Files.
