Written and maintained by CASRAI Editorial Board
Last updated
SPSS runs a Spearman rank correlation through the same Bivariate Correlations procedure used for Pearson’s r — it is a checkbox in the same dialog, not a separate menu. This guide covers the procedure, when to check Spearman instead of (or alongside) Pearson, how SPSS handles tied ranks, how to read the output, and the APA reporting line. For the underlying statistic itself — what a rank correlation measures and how it differs conceptually from Pearson’s r — see the correlation coefficient guide; for the parametric sibling procedure, see Pearson correlation in SPSS. This page covers the Spearman procedure specifically.
Running the Bivariate Correlations procedure for Spearman
Go to Analyze > Correlate > Bivariate. Move the variables into the Variables box — as with Pearson, entering more than two produces a full matrix of every pairwise correlation, not a single value.
- Correlation Coefficients: check Spearman and uncheck Pearson unless you deliberately want both coefficients printed side by side for comparison — a legitimate thing to do when you want to show a reviewer that a nonlinear or outlier-driven pattern changes the two numbers differently, but not a default.
- Test of Significance: Two-tailed is the default and correct unless the hypothesis specifies direction in advance, same rule as Pearson.
- Flag significant correlations: leave checked; it adds the same asterisk convention described below.
- Options: Missing Values is set here to Exclude cases pairwise or Exclude cases listwise — it works identically to the Pearson procedure (see that guide’s Options section for the full explanation), since both coefficients share one Missing Values setting when checked together in the same run.
SPSS computes rho by ranking each variable separately (ties are assigned the average of the tied ranks) and then applying the correlation formula to the ranks rather than the raw scores, with a tie-adjustment built in — you never rank the data yourself first.
When to check Spearman instead of Pearson
Check Spearman rather than Pearson when any of the following applies, and check both (see above) when you’re not sure which describes your data:
- The data are ordinal, not interval/ratio — ranks, Likert-type items treated as ordinal, or any variable where only the order of values is meaningful. Pearson’s r assumes interval or ratio measurement; Spearman only requires that both variables can be ranked.
- The relationship is monotonic but not linear — it consistently rises (or consistently falls) but not at a constant rate. Pearson’s r specifically measures straight-line association and will understate a strong monotonic relationship that curves; Spearman, working on ranks, is insensitive to the curve as long as the direction never reverses.
- One or both variables are heavily skewed, or the data contain outliers you can’t justify removing — because Spearman correlates ranks rather than raw values, a single extreme score has no more influence on rho than any other case once it’s been converted to a rank. This is the main practical reason researchers reach for Spearman on real (non-normal) data even when both variables are technically continuous.
- Sample size is small and you can’t reasonably assess whether Pearson’s normality-adjacent assumptions hold. Spearman’s assumptions are lighter (monotonicity, not linearity or bivariate normality), which makes it the more defensible default in a small, non-experimental sample.
None of this makes Spearman a safer all-purpose substitute for Pearson: when the relationship genuinely is linear and the data are interval/ratio, Pearson’s r is the more efficient and more interpretable estimate, and reporting Spearman instead invites a reviewer to ask why. Use the statistical test decision tree if you’re deciding between correlation and a different analysis entirely.
Reading the Correlations output table
With Spearman checked, SPSS prints the same row/column matrix layout as Pearson, but the label changes and the numbers underneath it come from a different calculation:
- Correlation Coefficient: rho, from −1 to +1. Same diagonal-is-always-1.000 convention as Pearson.
- Sig. (2-tailed): the p-value for the null hypothesis that the true rank correlation is zero — a significance test on rho, not on r, even though the row layout looks identical to the Pearson table.
- N: the number of cases used for that cell, subject to the same pairwise/listwise logic as Pearson.
If you checked both Pearson and Spearman in the same run, SPSS prints two separate matrices, one per coefficient — they are not merged into one table, so make sure you’re reading rho out of the Spearman block and not accidentally reporting the Pearson r from the block above or below it. The same significant-correlation asterisks (* p < .05, ** p < .01) and the same uncorrected-multiple-comparisons caveat described in the Pearson correlation in SPSS guide apply here without change — a k-variable Spearman matrix runs the same k(k−1)/2 independent tests, with no built-in correction, and the same guidance (Bonferroni on the comparisons you actually interpret, or an explicit exploratory-matrix disclosure) applies.
Writing it up in APA format
The standard reporting line is:
rs(df) = [value to two decimal places], p = [exact value, or p < .001 if SPSS rounds it to .000]
For example: rs(38) = .52, p = .001. As with Pearson, df = N − 2 (for a sample of 40 pairs, df = 38, not 40 — SPSS’s own N in the output table is not what goes in parentheses). Name the coefficient explicitly — “Spearman’s rho” or “Spearman’s rank-order correlation,” not the generic word “correlation” — since a reader needs to know which assumptions and which kind of relationship (monotonic vs. strictly linear) the reported number actually reflects. If you ran both coefficients for comparison, report both with a short line explaining why (e.g., an outlier or a curved-but-monotonic pattern that made the parametric and rank-based estimates diverge), rather than silently reporting whichever number is larger.
Frequently asked questions
Does Spearman require normally distributed data in SPSS?
No. Spearman’s rho has no normality assumption — it only requires that both variables can be meaningfully ranked and that the relationship between them is monotonic. That’s the main reason to prefer it over Pearson when a variable is skewed or contains outliers you can’t justify removing.
How does SPSS handle tied values when computing Spearman’s rho?
Tied scores are assigned the average of the ranks they would otherwise occupy (e.g., two tied third-place scores both get rank 2.5), and SPSS applies a tie-corrected version of the rank-correlation formula automatically — you don’t need to break ties by hand or choose a setting for it.
Can I run Pearson and Spearman together in the same SPSS procedure?
Yes — check both boxes under Correlation Coefficients in the Bivariate Correlations dialog. SPSS prints two separate matrices in the output, one per coefficient; they share the Missing Values (pairwise/listwise) setting but are otherwise independent calculations.
What’s the difference between Spearman’s rho and Kendall’s tau?
Both are rank-based, non-parametric measures of monotonic association and both are available as checkboxes in the same Bivariate Correlations dialog. Kendall’s tau is generally considered a better estimate of the population correlation with small samples or a large number of tied ranks, and its value is typically smaller in magnitude than rho for the same data; rho remains the more commonly reported of the two in most published research.
Should I report rho or r²-style “variance explained” for a Spearman correlation?
Report rho itself as the primary statistic. Squaring rho is sometimes seen in practice, but it does not carry the same “proportion of variance explained” interpretation that r² has for Pearson, since Spearman is a monotonic-association measure on ranks, not a linear-model coefficient — treat any squared-rho figure with more caution than r², or avoid it, in an SPSS write-up.
For the SPSS procedure covering another rank-based test, see the Mann-Whitney U test in SPSS guide. For more SPSS procedure walkthroughs, browse the research tools hub.
