In short
In APA 7 a p value is italicized, given to three decimals and written without a leading zero: p = .009, p = .039, p = .734. Report the exact value wherever you can and use an inequality only below the floor, p < .001. Never write p = .000, however firmly SPSS prints it: that .000 is a number rounded to three decimals, not a zero. By the same logic you do not write p = 1.000 either but p > .999, unless it is an adjusted p that the method itself truncates at 1. In a table you pick one convention — an exact p column or asterisks (*p < .05, **p < .01, ***p < .001) — and keep it identical across every table in the manuscript.
The p value is the number that appears most often in a psychology paper and the one that is formatted worst. Not because the rule is hard, but because almost nobody writes it: it gets copied out of the SPSS Sig. column with its own formatting and its own zeros, which is how whole tables of .000 reach peer review. Here is the exact form, the two impossible values and what to do in a table. The rest of the section is covered in how to report results in APA 7.
Italics, three decimals, no leading zero
The canonical form fits on one line: p italicized, three decimals, no zero before the decimal point, a space on each side of the sign. p = .009. p = .039. p = .734. That is the whole rule; everything below is the set of cases where one line is not enough.
The italics are not decoration. APA 7 italicizes statistical symbols written in Latin letters — p, t, F, M, SD, N, d — and keeps Greek letters roman, which is why alpha = .05 stands upright while p = .05 leans, even though both refer to the same threshold. The common mistake is not forgetting the italics, it is overshooting them: the equals sign, the digits, the comma and the brackets around the degrees of freedom all stay roman.
The leading zero is settled by one question rather than by a list. Can this number exceed 1? A mean can, so M = 3.42 keeps its zero. A p value is a probability and never reaches 1, so it goes bare: .039, not 0.039. The same rule strips the zero from r, R2, standardized betas, alpha, omega, eta2 and Cramér's V. R prints 0.0393 and you have to delete that zero by hand, one value at a time.
Two small points of style that survive into otherwise clean manuscripts. APA writes p value as two words with no hyphen when it is a noun, and hyphenates only in adjectival use. And SPSS's Sig. does not exist in a paper: in your table that column is headed p, italicized and unabbreviated, because Sig. is what a software company called it, not a statistical symbol.
p = .000 is not a p value, it is a rounding
SPSS rounds significance to three decimals. Anything below .0005 prints as .000, and that includes .0004, .00003 and .0000000001. A p value of exactly zero would mean your data are literally impossible under the null hypothesis, which no test with a continuous distribution ever returns. What you are looking at is not a zero. It is a small number badly dressed.
What you write is p < .001. It is the only inequality the manual asks for in so many words, and it is a floor: below .001 you stop. Recent SPSS versions now print <.001 in several tables and save you the step, but older output is still circulating, and R hands you <2e-16, which is not something you paste into a manuscript either.
If you are curious about the real number, double-click the cell in the SPSS pivot table and more decimals appear. You do not need it for the paper. The exception is fields that run thousands of tests at once — genome-wide analyses, cluster-level EEG — where the exponent genuinely carries information and the field convention wins; there you report 3.2 × 10−8 and say in the Method that you are using that notation.
It is worth understanding why this irritates reviewers so much. To them a .000 in a table is not a rounding slip, it is evidence that you copied output without reading it. And someone who copied without looking in Table 2 probably did the same in Table 3, so everything after that gets read with a magnifying glass. It is the cheapest tell in a manuscript.
p = 1.000 is out too, with one genuine exception
The rule above has a mirror image at the other end. A continuous test does not return a probability of exactly 1 either, so a 1.000 on screen is the rounded form of something like .99987. What you write is p > .999, on the same logic as the p < .001 at the bottom.
The exception is real and not merely theoretical: adjusted p values. Bonferroni multiplies each value by the number of comparisons and truncates the result at 1, because a probability above 1 means nothing. With six comparisons and an unadjusted p of .38 you get 2.28, which is cut back to exactly 1.000 by construction. Writing p > .999 there would invent a precision that does not exist, so label the column pbonf and state in the note that values are truncated at 1. Exact tests on small tables, Fisher's included, also return a true 1.000.
While we are at the large end: do not write ns or n.s. APA 7 wants the value because the value informs. With alpha at .05, p = .734 and p = .058 are both non-significant and they have nothing in common; the first says there is no signal there at all, the second says your study ran out of power. An ns erases that distinction, and with it the only useful reading of the result.
Exact wherever you can, and the rounding trap at .050
APA 7 asks for the exact value — two or three decimals, three in practice — because p < .05 hides the difference between .049 and .001, and that difference is enormous. Reporting the threshold instead of the number also stops anyone using your result in a meta-analysis or recomputing anything from it. The exact value costs nothing extra and says considerably more.
Here is the trap nobody warns you about. A p of .0498 prints as .050 at three decimals, and so does .0502. With alpha at .05 the first falls inside and the second outside, and the two look identical in your table. The decision is made on the unrounded value, never on what the screen shows you, and if your result lives right there, the worst thing you can do is hang the whole paper on the third decimal. Give the confidence interval and the effect size, and let the reader see the uncertainty instead of hiding it behind an asterisk.
A p value on its own is not a result. It travels with the test statistic and its degrees of freedom and with an effect size and its interval, which is what APA has asked for since the sixth edition and what any decent journal now takes for granted. If you have the sentence half written and want to see it with the formatting applied, the APA 7 formatter hands it back assembled.
One last thing about the number you are copying: the p your software prints is two-tailed unless the table says otherwise. If you ran a one-tailed test and halved the value, say so explicitly in the Method and justify why the direction was specified in advance. An undeclared one-tailed p is the quietest way to turn a .06 into a .03.
Where the p goes in the sentence
The order inside the sentence is fixed: what you found in words, then the statistic with its degrees of freedom, then the p, then the effect size with its interval. “Students in the intervention arm reported lower rumination than controls, t(122) = 2.64, p = .009, d = 0.47, 95% CI [0.12, 0.83].” Note that d keeps its leading zero, because it can exceed 1, and p does not.
When there are two degrees of freedom, both go inside the brackets, separated by a comma and set in roman: “F(3, 156) = 6.71, p < .001, partial eta2 = .11”. Chi-square is the odd one out because the N goes inside the brackets as well: “chi2(1, N = 184) = 0.05, p = .823”. That is also a neat illustration of the paragraph above, since 0.05 keeps its leading zero: a chi-square can exceed 1 without any difficulty.
The alpha level is declared once, in the analysis plan, and not repeated in every sentence of the Results. What does need retiring is the gray-zone vocabulary: marginally significant, a trend toward significance, approached significance. None of it is in the manual, and a reviewer reads it for what it is, a p = .068 being asked for a favor.
And the sentence underneath all of it, still misstated in published papers: p is not the probability that the null hypothesis is true, nor the probability that your finding will replicate. It is the probability of observing data at least as extreme as yours if the null were true, which is a different statement. There are more misreadings of the same size in the common errors when interpreting results.
Already drafted it? Paste your Results section into Review one section: it reads it the way a reviewer would and returns the first objection, attached to the sentence that causes it.
p values in tables: an exact column or asterisks, not both
In a table the p can go in two ways and both are APA. The first is a column of its own, with an italicized heading, values aligned on the decimal point and < .001 written inside the cell where it applies, never a dash or a blank. It is the usual choice in regression tables, where you already have one column per coefficient and another for the standard error.
The second is asterisks, and they have their own grammar. They are assigned in ascending order, one asterisk for the largest threshold: *p < .05, **p < .01, ***p < .001. The probability note comes last in the block of notes, after the general note and any specific ones, with a full stop after each threshold. The scheme must then be identical in every table of the manuscript: if three asterisks mean .001 in Table 2, they cannot mean .01 in Table 4.
What you do not do is use both. An exact p column with asterisks on top repeats the same information and makes the reader check that the two agree, which is your job, not theirs. The dagger for p < .10 is not APA either: that row of little crosses reads as a plea for clemency and a reviewer will strike it. If the effect is there, argue it with the interval, not with a new symbol.
If you corrected for multiple comparisons, the table has to say so in the heading and in the note. Label the column pFDR or pbonf — the subscript roman, since it is an abbreviation and not a statistic — and use the note to declare the method and how many tests it was applied over. The most transparent layout gives both columns, unadjusted and adjusted. For a correlation matrix, which is where asterisks pile up fastest, the correlation table generator builds it with the probability note already in place.
If your software does not give you the exact p value and you need to get it from the test statistic, the t, z, chi-square and F tables give the critical values and compute the exact p for your statistic.
Frequently asked questions
How do you report a p value in APA 7 style?
Italicized, to three decimals, with no leading zero and a space on each side of the sign: p = .009. Give the exact value wherever you can and use an inequality only below the floor, p < .001. It never travels alone: the test statistic with its degrees of freedom and an effect size with its confidence interval go with it.
Can you write p = .000 if that is what SPSS says?
No. SPSS rounds significance to three decimals, so anything below .0005 prints as .000. A p value of zero would mean your data are impossible under the null hypothesis, and no continuous test returns that. What you write is p < .001, both in the text and inside the table cell.
How many decimal places should a p value have?
APA 7 allows two or three, and three is what is used in practice. With two decimals, .012 and .006 both collapse to .01, and below .01 is exactly where that difference tends to matter. Below .001 you stop giving decimals altogether and write p < .001.
Is the p value italicized, and does it take a leading zero?
Italicized yes, leading zero no. APA 7 italicizes statistical symbols written in Latin letters and keeps Greek letters roman, so p leans and alpha stands upright. The leading zero belongs only to statistics that can exceed 1: M = 3.42 keeps it, while p = .039 does not, and neither do r, R2, beta or eta2.
What do the asterisks mean in an APA table?
They mark significance thresholds and are assigned in ascending order, one asterisk for the largest threshold: *p < .05, **p < .01, ***p < .001. The probability note goes last in the note block, after the general and specific notes, and the same scheme is kept across every table in the manuscript. If you already give a column of exact p values, the asterisks are redundant.
Can a p value be 1.000?
Not from a continuous test: you write p > .999, mirroring the p < .001 at the other end. Adjusted p values are the exception, because Bonferroni multiplies by the number of comparisons and truncates the result at 1, so that 1.000 is exact by construction. Label it pbonf and explain the truncation in the table note.
How many .000s are still sitting in your results table?
A reviewer does not read those as a rounding choice. They read them as output pasted in unread, and everything after that gets checked twice.
Run it through the paper reviewer →With p italicized, three decimals, no leading zero, exact wherever possible and p < .001 only below the floor, that part of the manuscript stops causing trouble. If your worry is not the formatting but which test was the right one and whether your data can carry it, that is what my statistical consulting is for.