Use Pearson correlation when both variables are continuous, roughly normally distributed, and related in a straight line; use Spearman correlation when your data are ordinal (ranks), skewed, have outliers, or the relationship is monotonic but curved rather than perfectly linear. If you've plotted your data and can't tell which one your thesis committee expects, you're not alone — picking the wrong coefficient is one of the most common corrections flagged in stats reviews.
Key Takeaways
- Pearson's r measures the strength of a linear relationship between two continuous, normally distributed variables.
- Spearman's rₛ (rho) measures the strength of a monotonic relationship by correlating the ranks, so it works for ordinal data, skewed distributions, and outliers.
- The deciding factors are data type, distribution shape, and outliers — Pearson needs normal continuous data with no extreme outliers; Spearman is the safe fallback when those assumptions fail.
- Both range from −1 to +1, where 0 means no relationship and the sign shows direction (positive or negative).
- When in doubt, run both — if they agree closely, report Pearson; if Spearman is noticeably higher, your relationship is likely non-linear or outlier-driven.
What is the difference between Pearson and Spearman correlation?
Pearson correlation (r) works on the raw values of your two variables and asks: as one goes up, does the other go up (or down) in a straight line? Spearman correlation (rₛ) first converts every value into a rank (1st, 2nd, 3rd…) and then runs a Pearson correlation on those ranks. That single step — ranking — is why Spearman doesn't care about the exact spacing of your numbers, only their order. This makes Spearman robust to skew and outliers that would distort Pearson.
Think of it this way: Pearson asks "is the relationship a straight line?" while Spearman asks "does one variable consistently increase (or decrease) as the other does, even if the line bends?"
When should I use Pearson correlation?
Use Pearson correlation when both variables are continuous (interval or ratio), approximately normally distributed, and the scatterplot shows a linear pattern with no severe outliers. Classic examples include height and weight, hours studied and exam score, or age and reaction time — measurements on a real scale where a straight-line assumption is reasonable.
Before you trust a Pearson r, check three things:
- Data type — both variables must be continuous, not categories or ranks.
- Linearity — plot the two variables; the cloud of points should follow a straight line, not a curve.
- Outliers and normality — a single extreme point can inflate or flip Pearson's r, so screen your scatterplot and distributions first.
If any of those fail, Pearson can badly misrepresent your relationship — which is exactly when Spearman earns its keep.
When should I use Spearman correlation?
Use Spearman correlation when at least one variable is ordinal, when your data are skewed or contain outliers, or when the relationship is monotonic but not straight-line linear. Because Spearman works on ranks, it's the go-to for Likert-scale responses, ranked preferences, small samples, and any variable with a long tail.
Spearman is also the safer choice when you simply aren't sure your assumptions hold. It answers a slightly more forgiving question — "do these move together in a consistent direction?" — which is often the real research question anyway. If you're weighing rank-based versus parametric methods more broadly, the same logic that separates the Mann-Whitney U test from the t-test applies here.
Pearson vs Spearman: side-by-side comparison
| Feature | Pearson (r) | Spearman (rₛ) |
|---|---|---|
| Measures | Linear relationship | Monotonic relationship |
| Data type | Continuous (interval/ratio) | Ordinal, or continuous |
| Distribution assumption | Approximately normal | None (distribution-free) |
| Sensitive to outliers? | Yes — highly | No — robust |
| Works on ranks? | No (raw values) | Yes (converts to ranks) |
| Best for | Height vs weight, test scores | Likert scales, skewed data |
| Range | −1 to +1 | −1 to +1 |
A worked example: study hours and exam scores
Say you survey 12 psychology students and record weekly study hours and their final exam percentage. The relationship looks strongly positive — but one student studied 40 hours (way above everyone else) and still scored moderately, creating an outlier.
Running Pearson, you get r = .62, p = .031 — a moderate, significant positive correlation. But that lone 40-hour outlier is dragging the straight-line fit around.
Running Spearman on the same data, you get rₛ = .81, p = .002 — a much stronger relationship. Why the gap? Because Spearman ranks the outlier as simply "the highest study time" rather than letting its extreme value distort the slope. The ranks reveal that study hours and scores climb together very consistently, even though the raw relationship isn't a clean straight line.
The takeaway: when Pearson and Spearman diverge this much, it's a signal — your data violate the linearity or outlier assumptions, and Spearman is the more honest summary. Here you'd report Spearman and explain why in your methods.
How do I report correlation in APA 7 format?
APA 7 wants the coefficient italicised, degrees of freedom in parentheses (for Pearson), the exact p value with no leading zero, and ideally a confidence interval. Two clean examples:
- Pearson: r(10) = .62, p = .031, 95% CI [.06, .88]
- Spearman: rₛ = .81, p = .002
Note the degrees of freedom for Pearson equal N − 2 (here, 12 − 2 = 10). Always report the direction and strength in words too: "There was a strong, positive monotonic relationship between study hours and exam scores, rₛ = .81, p = .002." Getting the italics and leading zeros right is a frequent source of reviewer corrections — a good reason to let a tool format it for you.
Which correlation should I actually run?
If you can honestly tick "continuous, normal, linear, no outliers," run Pearson. If any box is unchecked — ordinal data, skew, outliers, or a curved-but-consistent pattern — run Spearman. When you're unsure, run both and compare: close agreement means Pearson is safe; a big gap means Spearman is telling the truth about a non-linear or outlier-driven relationship.
This is exactly the kind of decision StatRyx makes for you. Upload your two variables, and StatRyx checks the distribution and outliers, recommends Pearson or Spearman based on your actual data, runs the correct test, and hands you the APA 7 write-up — coefficient, p value, CI, and a plain-language interpretation. No wrestling with SPSS menus or second-guessing your assumptions.
Unlike SPSS (which starts at roughly $99 per month for a personal licence) or code-heavy R, StatRyx is built for researchers who need the right answer without a statistics degree. It sits alongside tools like JASP and jamovi as a modern alternative — but it's the only one that also chooses the test and writes the results paragraph for you.
Stop calculating this by hand — run it free in StatRyx → Try StatRyx
Frequently Asked Questions
Can I use Spearman instead of Pearson to be safe?
Yes — Spearman