APA 7 figures: what goes above the image, inside it and below it

You paste the SPSS chart into Word, type "Figure 1: Results" underneath it and move on. Two months later the production team asks for the figure "at publication resolution" and Reviewer 2 wants to know what those vertical bars are, standard error or confidence interval. Both problems are ten minutes of work before submission and a fortnight of delay afterwards. Here are the rules for an APA 7 figure, what goes above the image, what goes inside it and what goes below, the resolution and color choices that survive a real journal, and how to decide when a figure beats a table. For the wider picture of the Results section, start with how to report results in APA 7.

In APA 7, anything that is not a table is a figure

The boundary is simpler than people expect: a table presents information in rows and columns, usually numbers, and a figure is everything else. A bar chart is a figure, and so is the path diagram of a structural equation model, the PRISMA flow diagram of a systematic review, the photograph of the stimuli you presented, a map, a drawing of your apparatus and a screenshot of the app you evaluated. Chapter 7 of the manual governs all of them with one set of rules, so you do not have to learn a separate format for each type of chart.

The two series are numbered independently and in order of first mention in the text: Table 1 and Figure 1 can sit on the same page without any conflict, and a figure mentioned earlier gets the lower number even if the layout eventually pushes it further down the page. That is why you always refer to a figure by number, "as Figure 2 shows", and never as "the figure below" or "the next chart": the journal moves images around when it typesets, and those phrases lose their referent. Write "Figure" in full, too, because APA does not use "Fig." in the text.

The rule almost everyone breaks is the other one: the same data must not appear in both a table and a figure. If you already have a table of means and standard deviations by group, a bar chart of those same means adds nothing and eats half a page. Pick one. Where the figure sits depends on where it is going: in a dissertation or a coursework paper you can embed it right after the paragraph that mentions it, or place all figures at the end after the references, while most journals want one file per figure with the captions listed separately. For how the whole section reads once it is assembled, see this worked example of a Results section.

The number and the title go above the image, on two lines, with no colon

The number. "Figure 1", in bold, flush left, on its own line and above the image. No italics, no full stop, arabic numerals. It is the first thing the reader sees and the first thing a copy editor checks when they run through your formatting.

The title. It goes on the next line, in italics, and it says what the reader is looking at, not what you did. "Anxiety Scores Before and After the Intervention by Group" is a title; "Chart Showing the ANOVA Results" is not, because it describes the analysis instead of the content. There is no colon between the number and the title: they are two separate lines, not one broken sentence. In English the title takes title case, with the principal words capitalized, and it stays brief enough to be read in one breath.

That "above" is the visible difference from biomedical journals, where the figure caption sits below the image as one running block of text combining title and explanation. If you are submitting to a nursing, public health or medical journal with a manuscript written in APA, that is one of the format changes you will be asked for. It touches no data at all: it moves the text and merges it into a paragraph.

The most common formatting mistake, and the easiest to avoid, is Word's automatic caption tool: "Figure 1: ...", in gray, italicized, sitting below the image, with a colon in it. That is four things wrong in a single line. Type the number and the title by hand as two ordinary paragraphs, one bold and one italic, and the problem disappears.

The legend sits inside, the note sits below, and that is where the error bars live

The legend is the key to your symbols and it belongs inside the borders of the figure, not underneath it. It says which color, line type or fill belongs to which group, and it uses the same typeface as the rest of the image. If you only have two series and there is room to write a label at the end of each line, label them directly and delete the legend: you save the reader one eye movement per data point they inspect, and that shows up in how much of the figure they actually understand.

The note sits below the image and has three tiers in a fixed order. The general note opens with "Note." in italics and explains whatever the whole figure needs: the N, what the error bars represent, what the abbreviations stand for and where the image comes from if it is not yours. Specific notes use superscript letters (a, b, c) for one panel or one element. The probability note closes with the asterisks: * p < .05. ** p < .01. If you have nothing to say at one of those tiers, that tier simply does not appear.

If your figure has error bars and the note does not say what they are, the query is coming. And fairly so: on identical data, standard error bars are about half the length of 95% confidence interval bars, so the same difference looks dramatic or looks like nothing depending on which you drew. A complete note reads like this: "Note. N = 148 (74 participants per group). Error bars represent 95% confidence intervals around the mean. The between-group difference was statistically significant, t(146) = 3.02, p = .003, d = 0.50." There is nothing left to ask.

If the figure is not yours, the copyright attribution also lives in the note, not in a word processor footnote: "Note. Adapted from [title of the work], by A. Author, 2023, Journal Name, 45(2), p. 118 (https://doi.org/...). CC BY 4.0." Use "From" if you reproduce it unchanged and "Adapted from" if you modify it, and add the source to the reference list as well. One caveat that trips people up: citing is not the same as having permission. If the original carries no open license, you have to ask the publisher before you reproduce it. With photographs of participants, the permission is theirs, in writing, obtained before anything else.

Resolution, size and type: why pasting from SPSS fails

When you copy a chart out of SPSS and paste it into Word, what travels is a bitmap at your screen resolution, around 96 dots per inch, with the viewer's gray background thrown in. On screen it looks acceptable. In the journal's PDF, shrunk to a single column of 85 mm, the axes go soft and those 8-point labels end up at five. It is one of the commonest reasons production bounces a paper back to you after it has already been accepted, and it always arrives at the worst moment, with a publication date already agreed.

What publishers ask for is fairly stable across the field: 300 dpi for photographs and halftones, 600 dpi for combination art that mixes photography and line work, and up to 1200 dpi for pure line drawings. Better still, send vector files, PDF, EPS or SVG, for any statistical chart, because a vector has no resolution to lose and scales without pixelating. A path diagram in vector format is readable at any size; the same diagram as a 96 dpi PNG is readable at none.

In practice: in SPSS, open the chart in the Chart Editor and use File > Export instead of copy and paste. In R, ggsave("figure1.pdf", width = 85, height = 70, units = "mm") for vector output, or the same call with dpi = 300 if the journal insists on PNG or TIFF. In Word, use Insert > Picture > From File, never Ctrl+V. And size the figure to what it will really occupy: one column is about 85 mm and full page width about 175 mm. Drawing an enormous chart and shrinking it inside Word does not raise the resolution, it only crushes the lettering.

Lettering inside the image should be sans serif between 8 and 14 points, and those points are measured at the final printed size, not on the giant version filling your laptop screen. The ten-second test: set the figure to the width it will be printed at, lean back about an arm's length and read the y axis label. If you cannot read it, and you already know what it says, nobody else will.

Accessible color: the figure has to survive grayscale

Roughly 8% of men and 0.5% of women of European ancestry have some form of red-green color vision deficiency. That is one man in twelve, which in a mixed psychology sample works out at somewhere between 2% and 4% of participants depending on how it splits by sex, and the same proportion among the people who will review you. And red against green is precisely the pair almost everyone reaches for when they color "control group" and "intervention group". APA 7 is blunt about it: color cannot be the only thing carrying the information, and the figure has to stay readable when it is printed in grayscale.

The fix is not to give up color, it is to encode everything twice. Three trajectories on a line chart are distinguished by color and also by line type (solid, dashed, dotted) and by marker (circle, triangle, square). Two series of bars are distinguished by color and by fill pattern. And wherever it fits, a label written directly on the element beats any legend. Once you have done that, print the figure in black and white and look again: if you can still tell what everything is, it is finished.

For the colors themselves, use palettes that have already been tested rather than improvising with the color picker. The Okabe-Ito set, eight colors designed to stay distinguishable with color vision deficiency, works for categorical variables, and viridis works for continuous scales such as the heatmap of a correlation matrix. What to avoid is the rainbow or jet scale, which invents boundaries the data do not have, and red against green for opposite meanings. If you use red to flag high risk, flag it with a pattern or a label as well.

One rule of hygiene that is not in the manual but saves revision rounds: if your figure has twelve colors, your figure has no message. Color is there to point at the comparison that matters, and everything else can be gray. A chart with two highlighted lines and eight in pale gray is understood at a glance; the same chart with ten different colors is deciphered by nobody.

Which chart for which data, and when a figure beats a table

The short map, by type of data. To compare groups, a bar chart with error bars is the weakest option available, because it hides the distribution: a box plot, a violin plot or a cloud of raw points with the mean overlaid say the same thing and also show the spread and the outliers. For a relationship between two continuous variables, use a scatterplot with the regression line and its confidence band (r = .28, p = .002, N = 118), because the plot shows what the coefficient hides, such as a curved relationship or the three cases holding the correlation up on their own (the full write-up is in how to report a correlation in APA 7).

For repeated measures, use lines with time on the x axis and one point per measurement occasion. For an interaction, use two lines: this is the case where a figure is close to compulsory, because "the effect of the intervention depended on baseline symptom severity, F(1, 204) = 10.74, p = .001, partial η² = .05" is understood instantly when you watch the two lines diverge, and is not understood in a paragraph however well you write it (the rest of the reporting is in how to report an ANOVA in APA 7). For a model with paths, a mediation or a full SEM, use a diagram with the standardized coefficients written on the arrows. For a systematic review, the PRISMA flow diagram; for a meta-analysis, the forest plot; to show you checked your assumptions, a Q-Q plot.

A figure wins when what you need to convey is a shape: an interaction, a trajectory, the skew of a distribution, a ceiling effect, a scatter that fans out to the right. It also wins when there are so many points that nobody would read them one by one. A table wins when the reader needs the exact value to reuse or reanalyze it: regression coefficients with their intervals, a correlation matrix, the sample characteristics. And there is a third case that is neither: two means and a t belong in a sentence of running text, and deserve neither a table nor a two-bar chart.

The decisive test takes ten seconds and you can run it yourself: cover the figure and read the paragraph. If the paragraph already says everything, the figure is redundant and an editor will tell you so. If something is missing that you can only get by looking, the figure stays. And when the answer comes out as a table, the APA table generator will format it to the manual without you fighting Word's borders.

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.

Frequently asked questions

Does the figure title go above or below in APA 7?

Above. The number ("Figure 1", in bold) takes its own line above the image, and the title, in italics, goes on the line immediately below it, still above the image. The only thing that goes underneath the figure is the note, which begins with "Note." in italics. Putting a full caption below the image, with title and explanation together, is the biomedical convention (Vancouver, AMA), not APA, so it is one of the format changes to make if you switch target journals.

How do I cite a figure taken from another paper in APA 7?

With a copyright attribution inside the figure note, not with a parenthetical citation in the text. The formula is "Note. From [title], by Author, year, Journal Name, volume(issue), page (DOI). License" if you reproduce it unchanged, and "Adapted from..." if you modify it, and the source also goes in the reference list. The part people forget: citing does not grant permission. If the original figure carries no open license (CC BY, public domain), you have to ask the publisher for authorization before reproducing it.

What is the difference between a table and a figure in APA 7?

A table presents information in rows and columns, usually numbers; a figure is any other visual representation: charts, flow diagrams, path diagrams, photographs, maps or screenshots. They are numbered in two independent series (Table 1, Table 2... and Figure 1, Figure 2...) and share nearly all their formatting rules, with two exceptions: tables carry horizontal rules and no vertical ones, and figures can carry a legend inside the frame. The same data should never appear in both.

What resolution do journal figures need?

The usual requirements are 300 dpi for photographs and halftones, 600 dpi for figures combining photography and line work, and up to 1200 dpi for line drawings. For statistical charts the best answer is to send vector files (PDF, EPS or SVG), which lose no quality when scaled. What never works is copying the chart out of your statistics package and pasting it into Word: that travels at your screen resolution, roughly 96 dpi, and comes out blurred in the final PDF even though it looked fine in your document.

Can APA 7 figures be in color?

Yes, as long as color is not the only way of telling the information apart and the figure still reads in grayscale. In practice that means adding a second code, line type, marker shape, fill pattern or a direct label, and using colorblind-safe palettes such as Okabe-Ito for categories or viridis for continuous scales. Check your journal's policy too: many publish color free of charge in the online version and bill for it in print.

Do I have to mention every figure in the text?

Yes, every one of them, and by number: "as Figure 3 shows". That order of first mention is exactly what sets the numbering. Avoid "the figure below" or "the following chart", because the journal moves images when it typesets and those references end up pointing nowhere. And make the mention say something: "the results are shown in Figure 3" adds nothing, whereas "the between-group difference held at all three follow-ups (Figure 3)" does.

With the number in bold above, the italic title on the line below it, the legend inside the frame and a note that says what the error bars represent, your figure already meets APA 7. If you are preparing a submission and would rather have someone look at the figures, the tables and the analysis behind them before the editor does, that is part of what I do in statistical consulting: checking the numbers, leaving figures and tables in the exact format the journal asks for, and helping you answer the reviewers when they arrive.