How to Report a Correlation in APA 7 (r, p and the Table)

That single line is everything a standalone correlation needs: coefficient, degrees of freedom (N − 2), p-value and, if your journal asks for it, the 95% confidence interval. Once you have several variables, the correct format is no longer the text but the correlation table, and there is one fully built and ready to copy further down. What follows are the templates for Pearson, Spearman and partial correlation, how that table is put together, and the formatting mistakes that keep coming back. The general guide for every other test is in how to report results in APA 7.

And once you are comparing more than two or three variables, don't build the matrix by hand in Word: the APA 7 correlation table generator builds it straight from your coefficients, with means, SD and significance asterisks already in place.

What a correlation report must include

For an individual correlation you report the coefficient (r for Pearson; rs for Spearman), the degrees of freedom in parentheses (which are N − 2), and the p-value. When there are many variables, the right move isn't to list them in the text but to present a correlation table with each variable's means and standard deviations.

Three details separate a clean report from one the reviewer sends back. First, the degrees of freedom are not N: they are N − 2, so with 100 cases you write r(98), not r(100). Second, APA 7 asks you to pair the estimate with its confidence interval whenever possible: r(98) = −.34, 95% CI [−.50, −.16], p < .001. Third, the reader has to know which coefficient you used and why, and that belongs in the Method section, not inferred from Results. If your variables are ordinal or fail Pearson's assumptions, the route is Spearman or Kendall, and the guide on reporting non-parametric tests in APA 7 covers it. And if what you are missing is the previous step, actually computing the coefficient, that is in the Pearson correlation tutorial.

Example in the text

"The relationship between trait anxiety and academic performance was negative and statistically significant, r(98) = −.34, p < .001." For Spearman: "rs(98) = .29, p = .004."

The most commonly missed formatting detail: the correlation coefficient has no leading zero before the decimal point, because it cannot exceed 1 (you write r = .34, not 0.34). The same applies to the p-value. The r symbol is italicised; the Spearman "s" subscript is not.

Copy-ready templates

Swap the numbers for yours and the sentence is done. Note that the minus sign, when there is one, sits right against the decimal point, with no zero in front:

• Pearson, significant: "The two variables were negatively and significantly correlated, r(98) = −.34, p < .001."
• Pearson, non-significant: "No association was found between the two variables, r(148) = .07, p = .393." A null result gets reported with the same detail as a significant one, with the exact p, never with a bare "n.s."
• With a confidence interval: "r(98) = −.34, 95% CI [−.50, −.16], p < .001."
• Spearman: "rs(98) = .29, p = .004."
• Kendall: "τ_b = .21, p = .006."
• Partial correlation: "Controlling for age, the association held, pr(97) = .28, p = .005."

The correlation table

This is the preferred format when comparing several variables. It's presented as a triangular matrix (only the lower or upper half, not both), with variables numbered in rows and columns, means and SD in the first columns, and significance asterisks with their key at the bottom (for example, *p < .05, **p < .01). APA format: horizontal lines at the top, below the headers and at the bottom; no vertical lines. When that matrix holds many items measuring the same construct, it is also the starting point for exploratory factor analysis.

APA 7 correlation table example

Here is what it looks like once it is built, with the four elements the journal expects to see: table number and title, means and standard deviations, triangular matrix and a note underneath.

Four things to notice in that table. The diagonal is left empty (or filled with dashes), never with ones, and the upper half is left empty too: it would be the same information twice. Means and SD do keep the leading zero (0.74), because they can exceed 1; coefficients do not. The note declares the N and which coefficient it is. And if you have missing data and the sample size shifts from one pair of variables to another, say so there instead of printing a single N that isn't true: "N ranged from 138 to 142 due to pairwise deletion."

Common mistakes

The most frequent: adding the leading zero (0.45 instead of .45), omitting the degrees of freedom, not stating whether it's Pearson or Spearman, filling the full matrix instead of the triangular one, or reporting dozens of correlations in the text when a table would be far more readable. And, as always in APA 7, don't confuse significance with magnitude: an r = .12 can be "significant" with a large N and still be trivial. If you have the matrix fresh out of your software and are unsure what belongs in the text and what in the table, copy it here and it tells you what to report for each coefficient.

Let me add the ones I see most in real manuscripts. Asterisks with no key underneath, so nobody knows whether one asterisk means .05 or .01. Reporting 45 correlations, presenting the three that came out at p = .04 as findings, and mentioning neither a multiple-comparison correction nor the fact that the analysis was exploratory. Writing that the association "was significant" without stating its direction, when the sign is precisely what matters. And automatically translating r = .30 into "a moderate correlation": the classic cut-offs are a convention, not a law, and the defensible move is to interpret the coefficient against what gets published in your own field, as I explain in the guide to interpreting effect sizes.

Before you submit: a badly built correlation table or an r with no degrees of freedom is exactly the kind of detail a reviewer flags on the first read, along with bigger issues you can no longer see because you have lived inside the manuscript for months. If yours is already written, run it through the Q1 Reviewer: a free Reviewer 2 style pre-review that tells you what they will pick on before the journal does.

Frequently asked questions

How do you report a correlation in APA 7?

Coefficient, degrees of freedom in parentheses and p-value, in that order: r(98) = −.34, p < .001. The degrees of freedom are N − 2, not N, so with 100 cases you write r(98). For Spearman it is rs(98) = .29, p = .004, and APA 7 also asks you to add the 95% confidence interval whenever you can.

Does r take a leading zero?

No. The correlation coefficient cannot exceed 1 in absolute value, so APA 7 asks you to drop the zero: r = .34 and r = −.34, never 0.34. The same rule applies to p, to factor loadings and to alpha. Means, standard deviations and any value that can go above 1 keep their zero.

How many decimals should a correlation have?

Two decimals for the coefficient and for means and SD; three for the p-value. When p is smaller than .001 you don't print the exact number, you write p < .001. And avoid rounding to .00: if the correlation is .004, say in the text that it is essentially null.

Pearson or Spearman?

Pearson assumes a linear relationship and reasonably continuous quantitative variables, and it is sensitive to outliers. Spearman works on ranks, handles ordinal variables, and copes far better with outliers and monotonic non-linear relationships. The decision comes from looking at the scatterplot, not just from a normality test, and it gets explained in the Method section.

What does an APA 7 correlation table look like?

A triangular matrix, lower or upper half but never both, with the variables numbered in rows and columns, means and SD in the first columns and significance asterisks with their key underneath. Horizontal rules at the top, under the headers and at the bottom, no vertical rules. The diagonal is left empty rather than filled with ones, and the note states the N and which coefficient it is.

Do I need to correct for multiple comparisons in a correlation table?

It depends on what you use it for. If the table is descriptive and sets the scene for your main model, present it and don't hang conclusions on stray asterisks. If you draw confirmatory hypotheses out of that matrix, then yes: apply a correction (Holm or Benjamini-Hochberg FDR are usually more sensible than Bonferroni) and state it in the table note.

Do I report an effect size for a correlation?

The coefficient already is an effect size, so you don't need to add another index. What does help is pairing it with the 95% confidence interval and, if you want to talk about shared variance, with r² interpreted cautiously.

That's your correlation properly reported. If you then need to move from correlation to a real model, regression, mediation or SEM, we look at it in my statistical consulting for theses and papers.