The notation is this: M = 24.30, SD = 5.12, 95% CI [22.40, 26.20]. Square brackets, not parentheses. A comma between the limits, no "to". The confidence level sits before "CI" and only the first time, if every interval in the manuscript is a 95% one. And no percent sign inside the brackets, even when the statistic itself is a percentage.
This article is about notation, not about what a confidence interval means. If what you need is to understand what a CI is telling you and what it is not, that is in interpreting confidence intervals, with the five reading errors that come up most.
Where APA 7 requires a confidence interval
The manual does not ask for it "where possible": it asks for it as the rule. In APA 7 a confidence interval is mandatory alongside:
Any effect size you report. A Cohen's d without an interval is incomplete, and it is one of the things most often flagged in review.
Regression coefficients, when the table carries them.
Mean differences when that difference is the main result.
Association statistics: correlations, odds ratios, relative risks.
Where it is not needed is in routine descriptive statistics of the sample. The mean age of participants does not need an interval; the mean of the dependent variable in each group, if the contrast is about it, usually does.
The writing rules, one by one
Always square brackets. 95% CI [22.40, 26.20]. Never (22.40 - 26.20) nor "between 22.40 and 26.20" inside the statistical result.
The level is declared once. If every interval in the manuscript is 95%, you can say so in Method — "all confidence intervals are 95%" — and then write only CI [22.40, 26.20]. If you mix levels, each one carries its own.
The limits take the decimals of the statistic. If the mean carries two decimals, the interval carries two. The interval is not rounded separately.
The leading zero follows the statistic's rule. A correlation cannot exceed 1, so it goes without: 95% CI [.21, .48]. A mean on a 0-63 scale can, so it keeps the zero: 95% CI [8.10, 12.40].
A negative limit carries its sign inside the bracket with no space: 95% CI [−0.12, 0.48]. Use the typographic minus, not a hyphen.
No units inside the brackets. Units go in the text or in the table header.
The decimal separator in APA 7 is the point, whatever your local convention. It is one of the differences that trips people up moving from a national journal to an international one.
Examples, ready to copy
Mean difference with a t test:
"The intervention group scored lower on anxiety (M = 18.70, SD = 4.88) than the control group (M = 24.30, SD = 5.12), t(58) = 4.31, p < .001, d = 1.12, 95% CI [0.57, 1.66]."
Note that the interval belongs to the effect size, not to the raw difference. If you want to report both, you have to say which is which: "the difference was 5.60 points, 95% CI [3.00, 8.20], d = 1.12, 95% CI [0.57, 1.66]". A lone interval sitting behind two statistics is ambiguous, and it is a guaranteed objection.
Correlation:
"Rumination correlated with depressive symptoms, r(118) = .34, p < .001, 95% CI [.17, .49]."
Odds ratio:
"The odds of relapse were higher in the group without follow-up, OR = 2.41, 95% CI [1.32, 4.40], p = .004."
With an OR, the interval is more informative than the p: if it contains 1, there is no effect. Write it in that order, interval before p, and the reader sees it without hunting.
Means and standard deviations: the notation most often missed
Three things recur in every manuscript I review:
M and SD go in italics. The symbol is italic; the number is not.
Inside a parenthesis they are separated by a comma, not a semicolon: (M = 24.30, SD = 5.12).
Decimals are the same for both and for the whole variable. If the mean carries two, the standard deviation carries two, and so does every group. A table with 24.3 in one row and 18.70 in another looks wrong, and it looks wrong immediately.
And an underlying rule APA 7 does spell out: decimals are chosen by the precision of the measure, not by habit. If your scale returns whole numbers from 0 to 63, two decimals on the mean is right and three is noise.
Reporting a bootstrap result
This is where most people come unstuck, because a bootstrap CI is not written like a parametric one: you have to say how it was built. Without that, the number is not replicable.
A complete bootstrap report carries four pieces:
The interval method. Percentile, BCa (bias-corrected and accelerated), or normal. They are not interchangeable, and BCa is the default expected in mediation.
The number of resamples. 5,000 is the accepted minimum today; 10,000 is standard. Writing "bootstrapping was used" without the number is like writing "an ANOVA was run".
The seed, if you want the result to be exactly reproducible. Not required by APA, but a journal with an open science policy will ask.
The interval, in the usual notation.
The sentence looks like this:
"The indirect effect of rumination on symptoms through avoidance was significant, ab = 0.18, SE = 0.06, 95% BCa CI [0.08, 0.31], with 10,000 resamples."
Two warnings that save a review round:
You do not report a p for a bootstrapped indirect effect. The decision is made with the interval: if it does not contain 0, the effect is significant. Putting a p next to it suggests the Sobel test was used, which is exactly what bootstrapping replaces.
A bootstrap interval need not be symmetric around the estimate, and that is fine: it is the point of the method. If yours comes out perfectly symmetric, check you did not request the "normal" interval by mistake.
If the bootstrap you need is for a mediation, the full table with direct, indirect and total effects is built by the APA 7 mediation table generator.
Mistakes that get flagged in review
Reporting an effect size without an interval. The most frequent one, and APA 7 asks for it explicitly.
Putting the interval in parentheses. Parentheses are for degrees of freedom; brackets are for the interval.
Writing "95% CI = [.17, .49]". The equals sign is surplus: it is written 95% CI [.17, .49].
Saying "there is a 95% probability the parameter lies in the interval". That is the Bayesian reading and it is not what a frequentist CI says. It is worked through in the interpretation article.
Bootstrap with no method and no resample count. The result stops being replicable, and that is a methods objection, not a formatting one.
Mixing confidence levels without saying so. A 90% hidden among 95% intervals reads as an attempt to make something significant.
The full notation — italics, decimals, the leading zero and the exact p — is chapter 2 of the book "Report Your Results in APA 7", and the bootstrap turns up applied in chapter 7 on the same study, with the dataset and the scripts to reproduce every interval.
Before you submit
Notation is one of the few things a reviewer checks in full, because it takes thirty seconds and says a lot about how the rest was done. If you want to run that check on the manuscript before they do, upload it to the Q1 paper reviewer: it is free and hands each objection back pinned to the sentence that triggers it.
References
American Psychological Association. (2020). Publication manual of the American Psychological Association (7th ed.).
Cumming, G. (2014). The new statistics: Why and how. Psychological Science, 25(1), 7-29.
Efron, B., & Tibshirani, R. J. (1993). An introduction to the bootstrap. Chapman & Hall.
Hayes, A. F., & Scharkow, M. (2013). The relative trustworthiness of inferential tests of the indirect effect in statistical mediation analysis. Psychological Science, 24(10), 1918-1927.