# How to Create a Figure Legend That Stands Alone: A Template By the end of this guide you will be able to write a figure legend that a reader can understand without opening your paper, hearing your talk, or asking you a question. You will have a reusable eight-slot template, two worked examples filled in end to end, and a short test that tells you whether the legend truly stands alone. Keep a few things on hand before you start: your figure files in their final export format (TIFF, PDF, PNG, or JPEG), the author guidelines for your target journal, conference, or course, your statistics output, and a plain text editor for drafting. If you are building research presentation slides rather than a manuscript, keep your slide software open too, because you will condense the same legend for the deck. The template below is free to copy and reuse. It works for a journal figure, a thesis, a poster, and a scientific presentation alike, and every slot in it exists because reviewers, editors, and readers keep asking for the same missing pieces. ## What a Standalone Legend Has to Do A legend stands alone when someone who sees only the figure and its caption can answer four questions. What question does this figure address? How were the data produced? What does every mark on the figure mean? What should I conclude? Anything less sends the reader hunting through the text. Anything more turns the legend into a second results section. Existing guidance says the same thing in fewer words. Among six principles for figures and tables in clinical research, the last reads simply: make the labels and legend clear and complete [1]. Peer reviewers hold legends to that standard. Published guidance for manuscript reviewers lists illustrative, well-labeled figures and detailed figure legends among the key items a reviewer must check before recommending publication [2]. The gap between that standard and everyday practice is measurable. In a methodological review of 35 early phase dose-finding oncology trials that reported patient-reported outcomes, 30 of the papers (85.7 percent) did not state clear objectives for those outcomes [3]. When the purpose of an analysis is vague, every figure and legend built on it inherits the vagueness. One caution belongs up front. A legend can clarify a good figure, but it cannot rescue a confusing one. The same clinical figure guidelines tell authors to include a graph only if it improves the reader's ability to understand the findings, and to aim for an immediate visual impression of the data [1]. If the figure itself needs work, fix that first, then write the legend. ## The Free Figure Legend Template for Papers and Research Presentations The template has eight slots. Each bracketed line is a decision, and the walkthrough below shows how to fill every one of them. Copy the block into your notes and reuse it for each figure you make. ``` Figure [number]. [Declarative title stating what the figure shows or found.] [Orientation sentence: name the figure type and map the panels.] [Methods slice: only the details needed to interpret THIS figure.] n = [number] [what n counts, e.g., biologically independent samples]. Data are from [number] independent experiments. Error bars represent [SD / SEM / 95% CI]. Groups were compared with [statistical test and correction]. Asterisks mark P values: *P < 0.05, **P < 0.01, ***P < 0.001. Abbreviations used in this figure: [ABBREVIATION, spelled-out term], [ABBREVIATION, spelled-out term]. [Images only: scale bar = [value] [unit], [stain or label], [magnification or objective], [instrument and key settings].] [Takeaway sentence: what the reader should conclude from this figure.] ``` ### The Eight Slots at a Glance | Slot | Question it answers | Example phrasing | |--|--|--| | 1. Number and title | What does the figure show? | "Figure 3. Drug Y halves invasion in vitro" | | 2. Orientation | What am I looking at? | "(A) Representative blots, (B) densitometry from three blots" | | 3. Methods slice | How were these data produced? | "Cells were treated for 24 hours, then lysed and probed" | | 4. Sample size | How much data sits behind each point? | "n = 5 mice per group" | | 5. Marks and abbreviations | What do the symbols mean? | "Filled circles, vehicle. Open circles, drug" | | 6. Statistics | How much confidence is justified? | "Error bars, SD. Two-way ANOVA with a post hoc test" | | 7. Image settings | Can I trust the image? | "Scale bar, 50 µm. Hematoxylin and eosin stain" | | 8. Takeaway | What should I conclude? | "Knockdown reduced invasion relative to control" | Two slots deserve a comment before the walkthrough. The title slot is declarative, meaning it states the content or the finding rather than labeling the chart type. The takeaway slot is the line most legends leave out, and it is the one that makes a figure work in a talk, where readers skim. ## Step-by-Step: Filling the Template ![Flow of steps for filling the figure legend template, from inventory through the declarative title to the statistics line and takeaway sentence.](/article-images/illustrations/figure-legend-template-research-presentation-eight-slots-order.webp) Work through the steps in order for each figure. The steps assume a data figure, and the notes flag where image-based figures differ. **1. Inventory the figure before you write.** List every panel, symbol, color, line style, and abbreviation the figure contains. Every item on that list must either appear in the legend or be self-explanatory from an axis label. A simple file habit keeps the inventory honest. Store each figure with a text file that shares its name: ``` fig1_dose-response.tiff fig1_dose-response_legend.txt ``` When the figure is revised, the paired text file is revised in the same sitting. Most legend errors in submitted manuscripts trace back to a figure that changed while its legend did not. **2. Write the declarative title.** Replace labels such as "Results of the migration assay" with a sentence that states the content or the finding: "Inhibitor X slows migration of treated epithelial cells." Declarative titles carry the message even for readers who skim only figure titles, which is how most people read a paper and how almost everyone reads a slide deck. **3. Write the orientation sentence.** Name the figure type and map the panels in one or two sentences: "(A) Representative phase-contrast images of scratch assays at 0 and 24 hours. (B) Wound closure at 24 hours as a percentage of the original scratch area." For a single-panel chart, one clause is enough: "Shown is a Kaplan-Meier curve, a plot of the proportion of subjects remaining event-free over time." **4. Write the methods slice.** Include only what a reader needs to interpret this figure: the model or sample, the treatment, the measurement, and any software choice that changes what the numbers mean. Reviewer guidance asks for methods described thoroughly enough that a competent colleague could repeat them, like a cookbook [2], but the full recipe belongs in your Methods section. The legend carries the slice that makes the figure readable: doses, time points, the definition of the endpoint, the stain by name. **5. Define n and count it honestly.** State what n counts. Biological replicates are independent units, such as separate animals, patients, or cultures. Technical replicates are repeated measurements of the same unit. If you ran an experiment 3 independent times with 2 technical replicates each time, your comparison runs on the 3 independent results: ``` 3 independent experiments, 2 technical replicates each = 6 measurements statistics computed on the 3 experiment-level values report n = 3 (biological replicates) and note the technical replicates separately ``` Reporting n = 6 in that situation overstates the evidence, and it is one of the first things a statistically trained reviewer checks. **6. Decode every mark.** Every symbol, color, line style, and shading pattern needs a definition: "Filled circles, vehicle-treated. Open circles, drug-treated. Shaded band, 95% confidence interval." If color carries meaning, choose palettes that stay distinguishable for readers with color vision deficiencies, a check that current figure assembly tools can run for you [4]. **7. Add the statistics line.** Three things belong here. First, what the error bars show: SD (standard deviation, the spread of the data), SEM (standard error of the mean, the precision of the estimated mean), or the 95% confidence interval (the range likely to contain the true value). Second, which test was used, for example an unpaired two-tailed t test or an ANOVA (analysis of variance) with a post hoc correction. Third, what the significance markers mean, for example asterisks tied to P < 0.05, P < 0.01, and P < 0.001. Spell the thresholds out once in the legend rather than making readers guess. **8. Add image specifics for micrographs and gels.** Any photographic image needs the stain or label, the scale bar value, and enough acquisition detail to judge what is shown. A primer on assessing mouse gross and histopathology images in the biomedical literature organizes the elements of an ideal pathology image and legend into image quality, image composition, and image interpretation, and warns that legends accompanying hematoxylin and eosin (H&E) or immunohistochemistry (IHC) images often contain inaccuracies [5]. A legend that names the stain, gives the scale, and states what a positive result looks like covers all three. **9. Write the takeaway sentence.** End with one sentence that states what the reader should conclude, phrased at the strength the data support: "Treated cultures closed the scratch more slowly than controls over the 24 hour imaging window." The conclusion must rest entirely on what the figure shows, which is the same standard reviewers apply to a study's overall conclusions [2]. **10. Run the stranger test, then format to spec.** Give the figure and legend to a colleague who does not know the project and ask the four questions from the start of this article. Any answer that requires your explanation marks a gap. Then open the author guidelines and confirm placement (some journals print legends under each figure, others collect them in a separate list), word limits, and file format. ### Worked Example 1: A Flow Diagram Legend Systematic reviews and scoping reviews usually open with a flow diagram of study selection, and its legend defines the search rather than the marks on the page. Here is a legend written for the flow diagram of a published scoping review of sex differences in cognitive decline and impairment in informatics research [6], using that review's real numbers: ``` Figure 1. Study selection for a scoping review of sex differences in cognitive decline and impairment in informatics research. The diagram maps the identification and screening of records. We searched Ovid and other databases (APA PsycInfo, EMB Reviews, and Embase) for peer-reviewed informatics journal articles and conference proceedings published from 2000 to 2025, following the PRISMA extension for scoping reviews (PRISMA, Preferred Reporting Items for Systematic Reviews and Meta-Analyses). Records were screened by title and abstract and then in full text against predefined inclusion criteria. 17 full articles met the inclusion criteria and form the evidence base for the review. ``` Note the shape. The title states what the diagram is, the middle defines the search a reader would need in order to judge coverage, and the final line fixes the count that everything downstream depends on. ### Worked Example 2: A Multi-Panel Experimental Figure Here is the same template filled in for a two-panel experimental figure. This is the expected output of the walkthrough. The data are invented for demonstration, so you can see slot placement without mistaking the numbers for a real study: ``` Figure 2. Inhibitor X slows migration of treated epithelial cells. (A) Representative phase-contrast images of scratch assays at 0 and 24 hours for vehicle control and treated cultures. (B) Wound closure at 24 hours, expressed as the percentage of the original scratch area closed. Confluent monolayers were scratched, rinsed, and treated with 10 micromolar inhibitor X or vehicle, and wound area was measured in ImageJ at both time points. n = 3 biologically independent cultures per group from 3 independent experiments. Each dot in (B) represents one culture. Error bars represent SD. Groups were compared with an unpaired two-tailed t test. Asterisks mark P < 0.05. Abbreviations: Con, vehicle control. TX, inhibitor X treated. Scale bar, 200 micrometers. Treated cultures closed the scratch more slowly than controls over the 24 hour imaging window. ``` Notice what the legend does not do. It does not propose a mechanism, it does not repeat the abstract, and it does not try to replace the exact numbers that belong in the results text, a separation the clinical figure guidelines draw explicitly [1]. ### Check the Result: The Stranger Test Run this test on every figure before submission or before the talk: ``` Hand the figure and its legend to a colleague outside your project. Ask them to answer, using only what they can see: 1. What question does this figure address? 2. How were the data produced? 3. What does every symbol, color, and abbreviation mean? 4. What should I conclude? Any answer that needs your explanation marks a gap in the legend. Fix the gap, then run the test again with a different reader. ``` A legend that passes with two different readers is ready for a journal, a conference, or a course. ## Adapting the Template to Different Figure Types ### Charts and Graphs For charts, the legend names the summary being plotted and the unit of analysis. Say whether bars show means or medians, whether a dot is one animal or one cell, and what the whiskers are. The clinical figure guidelines urge authors to think through how a graph can best convey the information rather than accepting the default chart types offered by statistical software [1], and the legend should reflect those deliberate choices. Keep exact numbers in the text, because graphs should not replace the reporting of key numbers there [1]. ### Microscopy and Histology Images Image-based figures carry extra duties. The pathology primer cited above frames an ideal image and legend around image quality, image composition, and image interpretation [5]. In practice, the legend names the stain (H&E for routine morphology, an antibody-based IHC label for a specific target), gives the scale bar, states the magnification or objective, and tells the reader what feature distinguishes a positive result from background staining. Because research scientists without pathology training sometimes assess these images on their own, an explicit legend reduces the risk of diagnostic error [5]. ### Flow Diagrams PRISMA flow diagrams and similar reporting diagrams carry their counts inside the boxes, so the legend defines the search rather than the marks. Name the databases, the date range, any language limits, and the screening steps, as in Worked Example 1. The same pattern applies to participant flow diagrams in clinical trials, where the legend defines the enrollment and allocation steps. ### Multi-Panel and Assembled Figures Multi-panel figures fail in a predictable way: the legend describes panels that no longer match the layout after a revision. Map every panel label (A, B, C, and onward) explicitly, and when panels share symbols, define them once and say they are shared. Assembly tools can carry part of the load. EasyFigAssembler, a web application for building multi-panel omics figures (large-scale molecular datasets such as proteomics or transcriptomics), supports nondestructive panel editing, automated grid layout, and export to TIFF, PDF, PNG, and JPEG [4]. Whatever tool you use, re-read the legend against the final assembled figure, not the draft. ## Figure Legends on Research Presentation Slides ![Comparison table of manuscript and slide legends across length, placement, methods detail, statistics and how each is read.](/article-images/illustrations/figure-legend-template-research-presentation-manuscript-vs-slide.webp) A scientific presentation is a different medium from a manuscript, and the legend has to change with it. Research presentation slides are read twice, once from the back of a room during the talk and once on a laptop afterward, because decks circulate as PDFs after seminars and conferences. Both readings count. Three adjustments cover most cases. 1. Condense. A slide legend runs two to three lines, not the manuscript's full paragraph. Keep the figure title, the definition of the marks, and the takeaway. Move the methods slice and the statistical detail into speaker notes or a backup slide you can pull up if a question comes. 2. Keep it readable. Text the audience has to squint at will not be read. Hold the legend at a size readable from a distance, and never shrink a full manuscript legend to wedge it under a slide figure. 3. Let the slide stand alone anyway. Your narration carries the detail during the talk, but the person reading the PDF next week has no narration. A condensed legend plus a declarative figure title preserves the message for that reader. The standard from journals carries over unchanged. The conclusion you state on a slide must rest entirely on what the figure shows, phrased at the strength the data support. ## Formatting, Tools, and Pre-Submission Checks Before you submit or present, work through four checks. 1. Venue rules. Open the author guidelines and confirm legend placement, word limits, and whether legends sit under each figure or in a collected list. The authors of the clinical figure guidelines note that their own principles can be set aside when breaking them produces a clearer figure [1], which is a useful reminder that venue judgment outranks any template. 2. File format and resolution. Export at the resolution the venue specifies, commonly around 300 dots per inch (dpi) for photographic images and higher for line art, in an accepted format such as TIFF, PDF, PNG, or JPEG [4]. 3. Color and accessibility. If color encodes data, use palettes that stay distinguishable under common color vision deficiencies. The Quality Advisor in EasyFigAssembler gives real-time feedback on effective resolution and color-blind-friendly palettes against journal requirements [4], and the same palettes work in any graphics program. 4. Legend-figure sync. Re-read the final exported figure against the final legend text, panel by panel. Mismatches almost always come from revising one without the other, which is why the paired-file habit from Step 1 pays off. Once a figure passes, save its legend to a personal library. Reusing tested slot phrasing across a paper keeps the voice consistent and cuts drafting time on the next manuscript or the next set of research presentation slides. ## Common Mistakes and Limitations These are the failures editors and reviewers flag most often, with the wording you may actually see in a query list. - Undefined abbreviations. Query: "Please define all abbreviations at first mention in the legend." Fix: add an abbreviations line, even for terms you consider obvious. - Unexplained error bars. Query: "Please state what the error bars represent." Fix: name SD, SEM, or the 95% confidence interval, and define the term at first use. - Ambiguous n. Query: "Please clarify what n refers to." Fix: state whether n counts animals, patients, cultures, or cells, and separate biological from technical replicates. - Legend-figure mismatch. Query: "The panel labels in the legend do not match the figure." Fix: re-read the exported figure against the legend after every revision. - A legend that repeats the results text. The legend interprets the figure, and the results section reports the numbers. When a legend starts duplicating whole paragraphs, cut it back to the eight slots. - A legend doing the Methods section's job. Full detail belongs in Methods. The legend carries only the slice needed to read the figure. - Missing image details. For micrographs, leaving out the stain, the scale bar, or the acquisition settings creates exactly the inaccuracies the pathology figure guidance warns about [5]. Two limitations of any template deserve a plain statement. First, a template cannot fix a figure that should not exist. If a graph does not improve the reader's understanding, cut it rather than caption it [1]. Second, venue rules override general advice. A conference abstract, a thesis, and a journal article apply different legend conventions, so treat the template as a default and the author guidelines as the final word. ## Frequently Asked Questions ### How long should a figure legend be? Long enough to pass the stranger test and no longer. A single-panel figure often needs three or four sentences, while a multi-panel figure can reasonably run past 200 words. Check your venue for a word limit before you draft. ### Where does the figure legend go in a manuscript versus a slide? In many journals the legend sits directly below the figure, but some collect all legends in a separate section, so the author guidelines decide. On research presentation slides, keep a condensed two-to-three line legend under the figure and move the rest to speaker notes. ### What does it mean for a legend to stand alone? A reader who sees only the figure and its legend can say what question the figure addresses, how the data were produced, what every mark means, and what to conclude. If any answer needs the paper's text or your narration, the legend is not done. ### Should the legend state the main finding? Yes. Put the finding in the declarative title or the final sentence, phrased at the strength the data support. A legend that only names the chart type wastes its most valuable line. ### How do I report n, error bars, and statistics in the legend? State what n counts, name the error bar measure (SD, SEM, or 95% confidence interval), name the statistical test, and spell out the significance markers and their P value thresholds. Define each abbreviation at first use. ### Do multi-panel figures need one legend or one per panel? One legend, with a panel-by-panel breakdown inside it. Label each panel (A), (B), (C) in the figure and mirror those labels in the legend so the reader can move between them without guessing. ### How do figure legends differ on research presentation slides? They get shorter. Keep the figure title, the symbol definitions, and the takeaway in two to three readable lines, and push methods and statistics into speaker notes or a backup slide, because the deck will be read without your narration. ### Is there a free figure legend template I can reuse? Yes. The eight-slot template in this article is free to copy, and the two worked examples show it filled in for a flow diagram and a multi-panel experimental figure. Adapt the slots to your venue's author guidelines. ## Sources 1. [Guidelines for Reporting of Figures and Tables for Clinical Research in Urology.](https://pubmed.ncbi.nlm.nih.gov/32542947/) 2. [Essentials of Scientific Manuscript Review.](https://pubmed.ncbi.nlm.nih.gov/39461744/) 3. [Statistical methods and data visualisation of patient-reported outcomes in early phase dose-finding oncology trials: a methodological review.](https://pubmed.ncbi.nlm.nih.gov/37781154/) 4. [EasyFigAssembler: Enhance Omics Data Storytelling through Effective Figure Assembly.](https://pubmed.ncbi.nlm.nih.gov/41781195/) 5. [A Primer for Research Scientists on Assessing Mouse Gross and Histopathology Images in the Biomedical Literature.](https://pubmed.ncbi.nlm.nih.gov/37712877/) 6. [Sex differences in cognitive decline and impairment: a scoping review in informatics literature.](https://pubmed.ncbi.nlm.nih.gov/41402904/)