How to Report an Exploratory Factor Analysis in APA 7th Edition

An exploratory factor analysis is reported in APA 7th edition as five blocks, always in this order: extraction method and rotation with a justification for the choice, sampling adequacy (KMO and Bartlett's test of sphericity), the criteria you used to decide the number of factors and what each one indicated, total variance explained, and a table with the pattern matrix and communalities. KMO goes before the variance explained, not after. And the table does not replace the text: both are required.

Below you will find the results paragraph ready to fill in with your own values, the loading table in APA 7 format with its note, and the five errors that get an EFA sent back. It does not matter which program you used: SPSS, R with the psych package, JASP and jamovi produce output that looks different, but what the results section needs is identical. It does not matter what field you work in either. The convention is the same for a health scale, a workplace climate questionnaire or a behavioral economics instrument. If you need the wording for the rest of your analyses, the general APA 7 results guide covers every test.

If you would rather skip the template and get the text already drafted, the EFA APA 7 report generator takes your values (loadings, KMO, Bartlett, variance explained) and returns the results paragraph and the table, ready to paste into the manuscript.

What APA 7 requires in an EFA write-up

The 7th edition APA manual does not dedicate a specific section to EFA, but publication norms in psychology and psychometrics journals have consolidated a clear standard. A complete write-up must include, in this order: (1) extraction method and rotation type, with a justification for choosing oblique or orthogonal rotation; (2) sampling adequacy indicators, meaning KMO and Bartlett's test of sphericity; (3) criteria used to decide the number of factors and what each indicated; (4) total variance explained by the final solution; and (5) a factor loading table with communalities. None of these is optional if you want to survive peer review.

A common mistake is omitting the rotation justification. Simply writing "Varimax rotation was applied" without explaining why will not satisfy any reviewer who knows psychometrics. Oblique rotation (Promax, Direct Oblimin) assumes factors can correlate, which is the norm when you measure constructs from the same domain, while orthogonal rotation (Varimax) forces independence. The choice should be grounded, even if briefly.

How to report KMO and Bartlett's test

Both go in a single sentence, before the extraction results: "The Kaiser-Meyer-Olkin measure indicated good sampling adequacy, KMO = .89, and Bartlett's test of sphericity was significant, χ²(28) = 1742.60, p < .001." Two decimals for the KMO and no leading zero, since it is bounded between 0 and 1. Bartlett's degrees of freedom are k(k − 1)/2, where k is the number of items: with eight items, 28; with ten, 45. If your output does not print them, they are worth computing, because a Bartlett's test with no degrees of freedom is an incomplete statistic.

Be careful with how you interpret both, because this is where write-ups tend to overclaim. The conventional reading of the KMO treats .50 as the floor and values from .80 upwards as good (Kaiser, 1974), but that global figure hides the per-item MSA, which is what tells you whether one particular item is dragging the matrix down; more and more journals ask for the minimum per-item value, and one sentence covers it if every item cleared the threshold. Bartlett's test, for its part, only rules out that your correlation matrix is an identity matrix, and with a large sample it comes out significant almost every time. Neither of the two validates your factor structure: they are a minimum condition for factoring to make sense at all, and they should be reported as such.

Results paragraph: template with real values

The following paragraph shows the standard format for a two-factor solution with eight items. Italicized values follow APA convention (statistical symbols) and values in brackets are placeholders for your own results:

"The [N] items were subjected to an exploratory factor analysis using principal axis factoring with [Promax / Direct Oblimin / Varimax] rotation. The Kaiser-Meyer-Olkin measure indicated [excellent / good / acceptable] sampling adequacy, KMO = [.xx], and Bartlett's test of sphericity, χ²([df]) = [value], p < .001, confirmed that inter-item correlations were sufficiently large for factor analysis. [Parallel analysis and Velicer's MAP test / Parallel analysis / The combination of the scree plot and parallel analysis] converged on [N] factors, which together accounted for [xx.x]% of the variance. [If cross-loading items exist: Item [x] showed cross-loadings on both factors (λ = [.xx] and [.xx], respectively) and was [removed from the analysis / retained based on theoretical grounds].] The factors, labeled [Factor 1 Name] and [Factor 2 Name], showed a [low / moderate / high] correlation, r = [.xx]."

Two style points reviewers check: statistical symbols (p, χ², r, h²) are always italicized; factor labels (Anxiety, Depression) and acronyms like KMO are not. If the rotation is orthogonal, drop the last sentence about factor correlations: by definition they are independent.

The loading table: structure and footnote

The table is the most visible element of the write-up and the one most often returned with revisions. Follow this structure:

• Number and title above the table: Table 1 (bold, no period) followed by a line with the title in italic title case. Example: Pattern Matrix for the Eight Items After Promax Rotation.
• Columns: item (abbreviated text if long), one column per factor, final column with communalities (h²). If rotation is oblique, the table shows the pattern matrix (pattern loadings), not the structure matrix.
• Value format: two decimal places with no leading zero (e.g., .78, not 0.78). Loadings below the suppression threshold (usually .30) can be omitted or left blank to aid readability.
• Bold the dominant loading for each item to guide the eye.
• General note below the table: state the extraction method, rotation type, and loading suppression threshold. Example: Note. Principal axis factoring with Promax rotation. Loadings below .30 are suppressed for clarity.

If the rotation is oblique, add the factor correlation matrix (Φ) below the table or report it in the text. Many journals require this explicitly.

Full example: table in APA 7 format

Five common errors reviewers send back

1. Reporting the structure matrix instead of the pattern matrix. With oblique rotation, the pattern matrix (configuration loadings) is what you present in the table. The structure matrix conflates the direct factor effect with indirect effects through factor correlations, making interpretation harder. SPSS labels both explicitly, so make sure you pick the right one.

2. Omitting the factor retention criterion. "Two factors were retained" is not enough. You must state which criteria you used and what each indicated. If parallel analysis suggested three factors and MAP suggested two, and you chose two, explain why.

3. Not reporting per-item KMO values. Many journals ask whether any item's MSA (measure of sampling adequacy) fell below .50 and what you did about it. If all items exceeded the threshold, one sentence is enough.

4. Confusing PCA with EFA. If you ran principal component analysis as if it were an EFA, a knowledgeable reviewer will catch it. PCA does not separate common from unique variance. If you already have results this way, be explicit about the method and its limitations, or re-run the analysis using principal axis factoring or unweighted least squares. The conceptual difference, and why reviewers care about it, is laid out in CFA versus PCA.

5. Leaving out communalities. Communalities (h²) show what proportion of each item's variance is accounted for by the extracted factors. They are required in a complete write-up. If any item has low communalities (< .30), comment on it.

This same EFA, worked from start to finish on a study with a reproducible dataset (240 cases, fixed seed) and R, SPSS and jamovi scripts, is chapter 9 of the book "Report Your Results in APA 7", together with scale reliability.

EFA vs. CFA: which goes in your results section?

A frequent question, especially from reviewers, is whether you should have run a confirmatory factor analysis (CFA) instead of an EFA. The short answer: it depends on your goal. EFA is appropriate when you are developing a new instrument or exploring the factor structure of an existing one in a new population. CFA is appropriate when you already have a hypothesized structure (from prior EFA or theory) and want to test whether it fits your data.

Many Q1 journals now expect both: EFA on a calibration subsample, then CFA on a holdout subsample (split-half cross-validation). If your sample is large enough (typically N > 400), this is the gold standard. If you report only EFA, acknowledge the limitation and state that CFA is needed in future work. When you get to that stage, the guide to reporting CFA in APA 7 follows the same logic as this one. If you need to run both analyses and report them cleanly, my questionnaire validation service covers the full pipeline.

Checklist before submitting the manuscript

Before hitting submit, verify that your results section includes: justified extraction method; rotation type and reason; global KMO and Bartlett's test with degrees of freedom and exact χ²; retention criteria (parallel analysis, MAP, scree) with their results; total variance explained; table with pattern matrix, bolded dominant loadings, communalities, and a note; inter-factor correlations if rotation is oblique; and mention of cross-loading or low-communality items.

Frequently asked questions

How do you report an exploratory factor analysis in APA 7?

As five elements in this order: extraction method and rotation with a justification for the choice; sampling adequacy through KMO and Bartlett's test of sphericity; the criteria used to decide the number of factors and what each one indicated; the total variance explained; and a loading table showing the pattern matrix with communalities. The table does not replace the text: both are required.

What KMO value is acceptable for an EFA?

The usual convention treats .50 as the floor: below that, factoring the matrix is hard to defend. From .60 the solution is usually considered workable, from .70 comfortable, and .80 or above is read as good (Kaiser, 1974). Report the overall value to two decimals with no leading zero, and check the per-item MSA as well: one item with an MSA below .50 can be the reason the global figure is mediocre.

How do I report Bartlett's test of sphericity in APA 7?

As a chi-square with its degrees of freedom and its p: χ²(28) = 1742.60, p < .001. It must be significant for the analysis to proceed, but note that with a large sample it is significant almost by default, so it is a minimum requirement rather than support for your solution. The degrees of freedom are k(k − 1)/2.

Which extraction method should I report in an EFA?

Principal axis factoring, unweighted least squares or maximum likelihood, and you have to name it. Maximum likelihood gives you fit indices and confidence intervals but assumes multivariate normality; principal axis factoring and ULS hold up better when that assumption fails, which is the common situation with Likert items. What you should not do is run a principal component analysis and call it an EFA: PCA does not separate common from unique variance, and a reviewer who knows psychometrics will catch it.

Oblique or orthogonal rotation in an EFA?

Oblique (Promax, Direct Oblimin) is the sensible default in psychology, because constructs from the same domain usually correlate and forcing them to be independent distorts the solution. Varimax imposes that independence and only makes sense if you are willing to defend it. With oblique rotation, what goes in the table is the pattern matrix, not the structure matrix, and you also report the factor correlation matrix.

How much variance explained is enough in an EFA?

There is no APA threshold, and the 50% or 60% figures that circulate are conventions borrowed from other traditions, not rules. In psychology, with principal axis factoring on Likert items, solutions between 40% and 60% are common and perfectly publishable. Report the total and the share of each factor, and above all do not use the percentage as a retention criterion: the number of factors is decided with parallel analysis or MAP, not by adding factors until the percentage looks respectable.

What loading counts as salient, and what counts as a cross-loading?

Conventionally .30 or .40, and whichever you choose has to be stated in the table note, because it determines what the reader sees. An item is usually treated as cross-loading when it loads above the threshold on two or more factors, or when its two highest loadings are close to each other. Flag those items and say what you did with them, whether you removed them or kept them on theoretical grounds, instead of letting them disappear silently from the table.

How many participants do I need for an exploratory factor analysis?

The rules of thumb you have heard, ten participants per item, a minimum of 300, are unreliable because the sample size you need depends on the data, not on a fixed ratio. What matters is the size of the communalities and how well each factor is determined: with high communalities (above .60) and three or four strong indicators per factor, samples in the 150 to 200 range can be enough, while with low communalities you may need several hundred (MacCallum et al., 1999). Report your N, report the communalities, and let the reader judge.

Before you submit: once the results section is written, run it through the Q1 Paper Reviewer. It is a free Reviewer 2 style pre-review that tells you what a reviewer would object to in your EFA (unjustified rotation, retention criteria that never appear, missing communalities, a table that does not match the text) before the journal says it and you lose three months.

If you want to review the methodological decisions that precede the write-up, that is, extraction, rotation and number of factors, the complete EFA guide covers each step in detail. And if you are preparing a full questionnaire validation (from EFA to CFA through reliability and measurement invariance), my statistical consulting service covers the whole process from data to accepted manuscript. See details on the questionnaire validation page.