How to build an APA 7 table: every rule, and the four tables you will actually use

Table formatting is where a manuscript quietly loses an afternoon, and it is not because the rules are hard. It is because your two tools pull the other way: SPSS hands you a full grid with lines everywhere, and Word offers a gallery of table styles built out of shading, banded rows and vertical borders. APA 7 wants none of that, and the whole rule set fits on one page. Here it is, one rule at a time, followed by the four tables that turn up in almost every psychology paper, with numbers that hold together so you can copy the shape and drop your own in. If you are not yet sure which figure belongs in which sentence, start with how to report results in APA 7.

The five parts of an APA table, in the order they appear

Every APA 7 table has five parts and they always sit in the same place. At the top, the number: “Table 1”, in bold, flush left, on a line of its own. Under it, the title, italicized, brief, in the same typeface as your text. Then the column headings, the body with the data and, at the foot, the notes. The title is a noun phrase, not a sentence: “Descriptive Statistics and Correlations Among Study Variables” is a title; “This table shows the descriptive statistics for the sample” is a caption that wandered in from a different style guide. One thing APA 7 tidied up is that figures now follow exactly this structure too, so learning one gets you both.

Every column takes a heading, including the leftmost one, which is the heading people forget. That first column has a name in the manual — the stub column — and its heading is usually “Variable”, “Predictor” or “Model”. Statistical symbols stay italicized in the headings (M, SD, t, p, n) and Greek letters do not (α, ω, β, χ²). When you group several columns under a shared label — “Control group” sitting above the M and SD columns — that label gets its own rule underneath, and that rule spans only the columns it covers, not the width of the table.

The body runs on consistency: the same number of decimals within a column, aligned on the decimal point, and no four-decimal values just because the software produced them. An empty cell means the data are not applicable; a dash means they were not obtained, and you say so in the general note. The test is simple: cover the article and see whether the table still makes sense on its own. If working out what “T1” means requires the Method section, that abbreviation belongs in the note. You can drop a point or two of font size to make a wide table fit, and the body can be single, one-and-a-half or double spaced, but do not switch typeface halfway through the document.

APA 7 lets you embed a table in the text right after the paragraph that calls it, or put them all after the references. That is a change from the sixth edition, which sent them to the back with no argument, and it is why so many university thesis templates still insist on it. Pick one option, keep it across the whole manuscript, and check what the journal wants before you lay anything out. One warning that saves whole tables: if it fits in a sentence, it is not a table. When your Table 3 is two means and a t, write “the intervention group scored higher, t(184) = 2.34, p = .020, d = 0.34” and delete it.

Rules: three horizontal, not one vertical

An APA table takes three rules and all of them are horizontal: one above the column headings, one directly below them, and one at the foot under the last row of data. Nothing else. There are no lines between the rows of the body, because white space already separates them. And there are no vertical rules at all, not to divide blocks and not “to make it easier to read”. If a table needs a vertical line before it makes sense, what it actually needs is fewer columns, or splitting into two tables.

The only line you add to those three is the spanner rule, which sits under a heading that covers several columns and extends only across them. In long tables some journals will also accept a rule separating blocks of rows — steps 1, 2 and 3 of a hierarchical regression, say — but it is not required and a blank row usually does the job.

In Word you get there by stripping every border out of the table and then adding back the three you want, not by picking a table style from the gallery: every style in that gallery ships with shading, vertical borders or banded rows, and none of those three things exists in APA. Paste from SPSS and the full grid comes with it, so it all has to go. Color is not in the system either, and it fails for a practical reason on top of the stylistic one: plenty of journals still print in grayscale, so a cell highlighted in yellow to flag significance simply disappears. Significance is marked in the probability note or in a p column.

The three tiers of notes, and the order they go in

Below the bottom rule come the notes, and APA has three types that are always written in the same order, each starting on its own line. The general note opens with an italicized “Note.”, full stop included, and carries whatever the whole table needs: what the abbreviations mean, what N is, which measure of variation you are reporting and, if the table is reproduced or adapted from another work, the copyright attribution and the permission behind it. It is the note people forget and the one that heads off the most reviewer queries.

Specific notes are flagged with superscript lowercase letters (^a, ^b, ^c) and attach to one column, one row or one cell. They are assigned in reading order, left to right and top to bottom, not in the order they occurred to you. The everyday case in psychology is an n that moves between measures: ^an = 93. ^bn = 89, because four people left that questionnaire blank.

The probability note comes last and holds the asterisks: * p < .05. ** p < .01. *** p < .001, always assigned in that order and meaning the same thing in every table of the paper. APA 7 would rather you gave the exact p value in its own column, and once you do the asterisks are redundant: do not run both, because a whole column is then repeating what sits next to it. If you keep the asterisks, remember that they live inside the table only. In running text you write the p value.

One small thing that marks a manuscript as unchecked: an asterisk in the body of a table that is never defined at the foot. It means nothing, and the reviewer has to guess whether it stands for p < .05, for p < .01, or for a footnote that was lost two drafts ago.

How to call a table out in the text

Tables are numbered in order of first mention in the text, not in the order you built them. Table 1 is the first one you name, even if it was the last one you made. You refer to it by number with a capital T — “in Table 2”, “the descriptive statistics are given in Table 1”. Never “the table above”, “the previous table” or “the table on the next page”: once the journal typesets the article the table can land two pages away, and your sentence becomes false without you touching a word.

The text does not repeat the table, it reads it. A well-written Results paragraph points at the pattern, pulls out one or two numbers and leaves the rest where it is: “The three subscales correlated in the expected direction (Table 2), with the strongest association between rumination and depressive symptoms, r = .38, p < .001.” If you find yourself transcribing the table row by row, either the table or the paragraph is redundant. There is a full section written this way in this APA 7 results section example.

Every table gets mentioned at least once, and a table that never appears in the text either goes or moves to supplementary material. Appendix tables are numbered separately with the appendix letter: Table A1, Table A2, Table B1. Supplementary tables are usually Table S1 onwards, though that one is house style rather than APA, and it is the place to be generous: the 18 × 18 correlation matrix that will never fit in the article belongs there.

The two you always need: Table 1 and the correlation matrix

Table 1 in nearly every psychology paper describes the sample. Its columns are “Variable”, “n”, “%”, “M” and “SD”, and its problem is that it mixes categorical and continuous variables in one grid: gender goes in n and %, age goes in M and SD. You fix that with a one-line general note rather than more columns: “Note. N = 186. Values are n (%) for categorical variables and M (SD) for continuous variables.” If you are reporting a trial and are tempted to add a column of p values comparing the arms at baseline, do not: CONSORT advises against testing baseline differences, because any imbalance in a randomized sample is by definition chance.

The correlation matrix is the table that goes wrong most often and is the quickest to put right. Fill one triangle only, usually the lower one, because the other is its mirror image; leave the diagonal empty or use it for each scale's reliability in parentheses, a widespread convention that has to be explained in the note. Means and standard deviations belong in the first two columns or the last two rows, never in a separate paragraph. Values carry no leading zero: r = .38, r = −.24, r = .21. If you are building it cell by cell, the APA correlation table generator produces it with the triangle, the decimals and the probability note already in place.

Four failures repeat in this table: four decimal places because the software gave four, both triangles filled in, a diagonal of “1.00” with nothing explaining it, and means kept in the prose while the correlations sit in the table, so the reader has to travel back and forth to reassemble something that belonged together. With N = 186 there is a fifth: reading the matrix by its asterisks alone. An r = .21 is significant at p = .004 and still accounts for under 5% of the variance.

The other two: the regression table and the model fit table

A multiple regression table almost always carries the same columns: “Predictor”, “B”, “SE”, “β”, “t”, “p” and the 95% confidence interval for B, which APA 7 asks for whenever you can give it. A real row reads: rumination, B = 0.42, SE = 0.11, β = .27, t(181) = 3.82, p < .001, 95% CI [0.20, 0.64]. Watch the leading zeros, because that is where most tables slip: B and both interval limits keep theirs, since they can exceed 1, while β and p never do. Overall fit is not another body row: it goes in the general note — “R² = .31, F(4, 181) = 20.33, p < .001” — or in a clearly separated summary row. What to report and what to leave out is covered in how to report a regression in APA 7.

A fit table for a confirmatory factor analysis or a structural equation model compares models, so its stub column is “Model” and each row is one of them: χ², df, CFI, TLI, RMSEA with its interval, SRMR and, for nested models, ΔCFI. A three-factor solution would read χ²(120, N = 186) = 178.60, p < .001, CFI = .946, TLI = .939, RMSEA = .051, 90% CI [.036, .065], SRMR = .052. Here the general note is effectively compulsory, because nobody is obliged to remember what TLI and SRMR stand for, and none of those indices takes a leading zero: they are .946 and .051, not 0.946 and 0.051.

One principle orders all four tables: a table answers a question. If you are pasting the entire output with its sums of squares, its mean squares and three empty columns, that is not a table, it is a data dump. Strip out everything that is neither cited in the text nor needed to reproduce the analysis, and if you cannot bear to lose it, send it to supplementary material. The universal APA table generator does that cleaning for you: paste the output and it returns the table with the rules, the decimals and the notes where they belong.

Already drafted it? Paste your Results section into Review one section and you get the first objection a reviewer would raise, quoting the exact sentence from your text. Free, no sign-up.

Frequently asked questions

How do you format a table title in APA 7?

On two lines, above the table. First the number, “Table 1”, in bold and flush left, on a line of its own. Underneath it the title in italics, not bold, in title case, brief and with no full stop: “Descriptive Statistics for Study Variables”. It is a noun phrase, so no verb and no “Table showing…”. Figures follow the same structure in APA 7, so the rule transfers.

Can APA tables have vertical lines?

No, never. An APA table takes three horizontal rules: above the column headings, below the headings and at the foot under the last row. The only line you add to those is a spanner rule beneath a heading that covers several columns, and it extends only across those columns. There are no lines between body rows either. If a table seems to need a vertical rule, it has too many columns and should be split.

How do I convert an SPSS table into APA format?

Three steps, none of them automatic. First, delete every border and add the three horizontal rules, because SPSS pastes its full grid. Second, rewrite the headings: “Sig.” becomes p, “Std. Error” becomes SE, “Beta” becomes β, and shouty variable names become readable ones. Third, fix the numbers: two decimals, no leading zero on p, r and β, and delete any column you will not cite. Then add the general note with the abbreviations and N.

How do you refer to a table in text in APA 7?

By its number with a capital T: “in Table 2” or “results are presented in Table 3”. Tables are numbered in order of first mention, so Table 1 is the first one you name even if it was the last one you built. Do not write “the table above”, “the previous table” or “the table overleaf”, because typesetting moves tables. Every table must be called out at least once in the text.

What do the asterisks in an APA table mean?

They are the probability note, defined in the last line of the notes at the foot of the table: * p < .05. ** p < .01. *** p < .001. They are always assigned in that order and mean the same thing in every table of the paper. APA 7 prefers exact p values in their own column, in which case the asterisks are redundant. An asterisk with no note defining it means nothing at all.

How many decimal places should an APA table have?

Two for most values, consistent within each column, with exact p values to two or three decimals unless they fall below .001, where you write p < .001. Drop the leading zero on any statistic that cannot exceed 1: p = .020, r = .38, β = .27, α = .89, CFI = .946. Keep it on statistics that can: M = 3.42, SD = 0.81, B = 0.42, d = 0.34.

If you would rather have someone go through the tables and the Results section before the editor does, that is part of what I do in statistical consulting: putting every table in the exact format your target journal asks for, checking that the numbers in the text and the numbers in the table are the same, and helping you answer the reviewers when they arrive.