APA 7 Table Notes and Significance Asterisks

Nobody reads the notes under a table until an editor sends them back marked up, and then it turns out the problem was never the analysis: it was three lines of small print with undeclared asterisks, an abbreviation only the author understands, and a footnote marker set in a symbol APA reserves for something else. This is the full set of rules — the three note types, the order they go in, what each asterisk means, and why more and more journals would rather you dropped them. For the wider picture of the whole section, start with how to report results in APA 7.

Three types of note, in a fixed order

APA 7 recognizes three kinds of note beneath a table and always places them in the same sequence: the general note first, covering the table as a whole; then specific notes, which point at a column, a row or a single cell; and the probability note last, explaining the asterisks. They sit below the bottom rule of the table, flush left, with no indent, and each type begins on a new line.

Not every table needs all three. A table of means and standard deviations with no tests carries no probability note; a table with no unusual abbreviations may need no general note. What never changes is the order of the ones you do include, and the fact that each type gets its own line. Running all three together into one block of small print is the most common mistake and the easiest to avoid.

Within a single type, though, the notes run on in the same line, separated by full stops. Three specific notes take up one line, not three. The superscript letters follow the order in which they appear in the table, reading left to right and top to bottom.

It is worth being precise about placement, because journal production checks it before anything else: the table number and title go above the table (Table 1 in bold on its own line, the title italicized beneath it) and the notes go below. A note placed above the table, or converted into a document footnote with a superscript numeral, comes straight back for correction.

The general note: the table's glossary

It opens with the word Note. in italics, full stop included, and everything after that is set in roman. This is where you explain what a reader cannot work out from the grid itself: what each abbreviation stands for, what units the figures are in, what the numbers in brackets are, how a binary variable was coded, or why some cells are empty.

Here is one from a correlation table in a study of undergraduate sleep and burnout: “Note. N = 186. AB = academic burnout; CI = confidence interval. Coefficients below the diagonal are for first-year students and those above the diagonal for final-year students. Gender was coded 0 = female, 1 = male.” Four things that will not fit inside the table and without which the table cannot be read on its own.

That is the standard to hold yourself to: a table has to stand alone. Even if you defined “AB” in the Method section, you define it again here, because the table is the first thing a reviewer looks at and the only part that gets reproduced when somebody drops your result into a slide. Repeating the definition is not redundancy, it is the rule.

If the table is reproduced or adapted from another source, the copyright attribution goes at the end of the general note — not in a separate note and not in the title. That is where the permissions desk at the publisher looks for it, and putting it anywhere else is the fastest route to being asked for the permission letter twice.

Specific notes: superscript letters, never asterisks

When the clarification applies to one column, one row or one cell rather than the whole table, it takes a superscript lowercase letter. The letter goes in the table, attached to the element it refers to, and the note below opens with the same superscript letter. Lettering starts at a in every table: Table 3 does not continue Table 2's sequence, it starts again.

A line of specific notes looks like this: “^an = 94. ^bn = 92. ^cThis subscale is reverse scored; a higher score indicates better sleep quality.” They run on, separated by full stops, on a single line and in alphabetical order, which is the reading order of the table.

The warning that saves a round of copy-editing: do not use asterisks for this. The asterisk belongs to probability, and if it also marks “estimated value” or “imputed cell”, the specific note and the probability note are competing for the same symbol and the reader cannot tell which one applies. Superscript numerals are out too, because they collide with exponents and with your word processor's own footnote markers.

The most frequent use of a specific note is declaring a different n for one column. If a subgroup lost participants to missing data, that is where you say so, not in a sentence buried in the text that a reviewer has to hunt for. It is one of the things most often missing at review and it costs a dozen words.

The asterisks: what each one means and how to set them

The convention is the familiar one: one asterisk for p < .05, two for p < .01 and three for p < .001. Fewer asterisks means a larger p value. The probability note is set on its own line, last in the block, like this: “*p < .05. **p < .01. ***p < .001.” No space between the asterisk and the p, the p italicized, no leading zero on the values, and a full stop after each entry.

Inside the cell the asterisk sits flush against the value: r = −.38*** for a correlation at p < .001, .21** for one at p = .004, −.16* for one at p = .029 and a bare −.11 for one at p = .135, all on N = 186. Two rules get broken constantly here. The note must be declared in every table that uses asterisks, because a table does not inherit the previous table's footer; and the same number of asterisks must mean the same alpha throughout the manuscript. If two asterisks are .01 in Table 2, they cannot be .001 in Table 4.

Do not declare levels you never use. If none of your coefficients reaches .001, the note has two entries rather than three. And if you mix one-tailed and two-tailed tests in one table, APA 7 asks you to keep the asterisks for the two-tailed values and give the one-tailed ones a different symbol, such as a dagger, declared in the same note. What almost no journal will accept any more is a dagger for the .05 < p < .10 band labeled “marginally significant”: that threshold belongs to no convention, it always appears after the data have been seen, and it is a red flag to any trained reader.

Two software traps worth naming. SPSS flags only two levels in its correlation matrix, one asterisk for .05 and two for .01, so if you copy its asterisks but write a note declaring three, the table is lying: no three-asterisk cell can exist in it. R does the opposite — the signif. codes legend adds a full stop for p < .1 that is no part of APA style and slips into the tables of anyone who pastes the output wholesale.

Why journals increasingly want the exact p value

An asterisk is not a number, it is a category, and categorizing costs information. On N = 186, a correlation of .15 (p = .041) and one of .18 (p = .014) carry exactly the same single asterisk, although one is scraping the threshold and the other is not. Worse, a coefficient of .14 comes out at p = .057 and gets nothing at all, even though it sits far closer to .05 than to the .50 a reader might imagine behind a bare cell.

The second reason is that a table of asterisks cannot be checked. Anyone who wants to reanalyze your result, feed it into a meta-analysis, or simply verify that the statistic, its degrees of freedom and the p value are mutually consistent needs all three numbers. Nuijten and colleagues (2016) ran an automated checker across tens of thousands of psychology articles and found at least one inconsistency between a reported test statistic and its p value in close to half of the papers that reported significance tests. With asterisks, that check cannot even be attempted.

The third reason is the deep one. The asterisk is the typographic form of a yes-or-no verdict, and the field has spent a decade arguing against exactly that: the American Statistical Association's statement on p values, written by Wasserstein and Lazar (2016), asks explicitly that evidence not be collapsed into a threshold. A table that only says “passed” or “did not pass” pushes the reader into the reading the discipline is trying to dismantle.

There is a practical argument that closes the case. If the table already carries confidence intervals, the asterisk is redundant. A 95% CI that excludes zero says what the asterisk says and also says how much, and how precisely. APA's own reporting standards (JARS-Quant) ask for the exact test result alongside an effect size and its interval, so the table most editors now want simply has no room left for a row of stars.

What the table looks like with exact p values

In a regression table it is trivial: the p value takes its own column and the probability note disappears. One row reads: sleep quality, B = −0.42, SE = 0.11, β = −.31, t = −3.82, p < .001, 95% CI [−0.64, −0.20]; and the next, screen time before bed, B = 0.20, SE = 0.09, β = .16, t = 2.22, p = .028, 95% CI [0.02, 0.38]. Note the formatting detail: B takes a leading zero because it can exceed 1, β and p do not, because they cannot. The rest of the table conventions are in how to report a regression in APA 7.

In a correlation matrix the constraint is space, and there are two clean ways out. The first is the double triangular matrix: coefficients above the diagonal, exact p values below it, and a general note saying which half is which. The second is to replace the asterisks with the confidence interval in brackets under each coefficient, which is what several clinical psychology journals now require. If you are assembling it by hand, the APA correlation table generator hands it back already formatted, and the analysis itself is covered in how to report a correlation in APA 7.

With multiple comparisons the asterisk becomes actively misleading, because it does not say whether the value behind it is raw or adjusted. If you applied Bonferroni, Holm or a false discovery rate correction, the honest layout is a column of adjusted p values, or the raw p plus a general note naming the method and the number of contrasts. A lone asterisk in that table means nothing anyone can verify.

Before it goes out, run one pass over the typography: italics where they belong, Greek letters upright, leading zeros only on statistics that can exceed 1, and p values to two or three decimals except those written as < .001. If you would rather not check it by eye, the APA 7 formatter runs that check on whatever text you paste into it.

Already drafted it? Paste your Results section into Review one section before it goes to your supervisor or the journal. You get the first thing a reviewer would object to, with your own sentence quoted.

If you would rather not set the notes and asterisks by hand, the APA table generator builds the table with its general and probability notes.

Frequently asked questions

What do the asterisks mean in an APA table?

They mark the significance level: one asterisk is p < .05, two are p < .01 and three are p < .001. Fewer asterisks means a larger p value. The meaning has to be declared in the probability note at the foot of that same table, and it must stay the same across every table in the manuscript.

What are the three types of table note in APA 7?

The general note, which opens with an italicized Note. and applies to the whole table; specific notes, flagged with superscript lowercase letters, which explain one column, row or cell; and the probability note, which defines the asterisks. They go in that order below the table, and each type starts on a new line.

How do you write a probability note in APA 7?

On its own line, last in the block, formatted like this: *p < .05. **p < .01. ***p < .001. No space between the asterisk and the p, the p italicized, no leading zero on the values, and a full stop after each entry. Only declare the levels the table actually uses.

Should you report asterisks or exact p values in tables?

APA 7 permits asterisks with a probability note, but many journals now ask for a column of exact p values to two or three decimals, with < .001 where appropriate. The reason is that an asterisk hides the distance to the threshold (p = .041 and p = .014 carry the same one) and makes it impossible to check the statistic, its degrees of freedom and the p value against each other. Check the journal's own instructions before you decide.

Can you use a dagger for a marginally significant p value?

It is a practice more and more journals reject. The band between .05 and .10 gets labeled “marginally significant” only after the data have been seen and corresponds to no established convention. APA does allow a dagger or another symbol to distinguish one-tailed from two-tailed tests within the same table, as long as it is declared in the note.

Do you have to repeat the asterisk note in every table?

Yes. Each table is a standalone unit and inherits nothing from the one before it, so every table using asterisks carries its own probability note. The same applies to the superscript letters of specific notes, which start again at a in each table.

Table notes are the kind of thing that takes ten minutes to fix once somebody tells you where to look, and costs a full round of review when nobody does. If you are closing the results section of a thesis or a paper and want the tables to land right first time, our statistical consultancy goes through them with you, one at a time, with the numbers and the target journal in front of us.