Installs into .claude/skills of the current project.
Are you the author of Figure Captions?
Add the live security badge to your README. It updates with every re-scan.
[](https://www.skillsdirectory.com/skills/aicodedecode-figure-captions)
---
name: figure-captions
description: Writing figure captions that make figures self-contained — what to include, ordering, and statistical reporting.
category: scientific
---
## Overview
A figure plus its caption should be understandable without reading the
paper — many readers see only these. This skill covers the caption's
job (identify, describe, explain, interpret-with-restraint), the
standard information each caption type needs (n, statistics, symbols,
scale), and the editing that keeps captions complete but tight.
## When to use
- Writing captions for a paper, thesis, poster, or presentation
- Making figures self-contained for skimmers and figure-browsers
- Reporting statistics, sample sizes, and methods in captions correctly
- Standardizing caption style across a multi-author paper
- Fixing captions that merely restate the title or describe the obvious
## Core concepts
- **The self-containment test:** cover the figure, read only the caption — can you state what was measured, how, in what, and what the result was? If not, the caption fails.
- **Caption anatomy:** (1) title phrase — what the figure shows; (2) method summary — what was done to produce it; (3) symbol/key definitions — every color, symbol, line, and abbreviation; (4) statistics — n, tests, p-values, intervals; (5) takeaway — one sentence on what it means (not a discussion, a pointer).
- **Define everything:** every abbreviation, symbol, color, error-bar type, and line style appearing in the figure — the caption is the legend's backup and often its replacement.
- **Statistics belong in captions:** sample sizes per group, the test used, exact p-values (not just stars — though stars are fine with a defined key), what error bars show (SD? SE? 95% CI? — always state).
- **Brevity with completeness:** captions are not methods sections — summarize the method in a clause, cite the full protocol; but never omit what a reader needs to interpret the panel.
- **Panel referencing:** multi-panel figures need (a), (b), (c) labels in both figure and caption, described in order — don't make readers hunt.
- **Caption length calibration:** one sentence for simple schematics, a full paragraph for multi-panel data figures — length should match information density, not author laziness or verbosity.
- **Statistical reporting standards:** exact p-values, effect sizes with intervals, test names, n per group, multiple-comparison corrections — the caption is where methods meet results; skimping here is the most common caption failure.
- **Accessibility:** captions are read by screen readers — write them as coherent prose describing the figure's content and message, not as telegraphic fragments.
## Practical workflow
### 1. Draft per panel, then assemble
1. For each panel write: what is plotted (axes, units), the experimental condition, the key observation with numbers.
2. Assemble in panel order with (a), (b), (c) markers; check the caption's order matches the figure's layout.
3. Add the shared information once (sample sizes, statistics, abbreviations) rather than repeating per panel.
### 2. Include the non-negotiables
- **n:** per group/condition, and what n counts (animals? cells? technical replicates?) — "n = 6" is ambiguous; "n = 6 mice per group" is not.
- **Statistics:** test name, exact p or the star key, correction for multiple comparisons, what the bars/bands show.
- **Scale:** scale bars on images (with length), axis units everywhere, colorbar units on heatmaps.
- **Replicates:** how many independent experiments; whether points are biological or technical replicates.
### 3. Write the takeaway sentence
1. End with one interpretive sentence: what does this figure establish? ("Together, these data indicate X mediates Y.")
2. Keep it to what the figure shows — save the broader interpretation for the Discussion.
3. Make sure the takeaway matches the Results text — caption and body must agree.
### 4. Edit for consistency
1. Standardize across all figures: "mean ± SD" vs "mean ± SEM" (pick one, define everywhere), p-value formatting, abbreviation style.
2. Cut redundancy with the main text — the caption summarizes, the text elaborates; they shouldn't be identical paragraphs.
3. Length check: if a caption exceeds ~200 words, consider moving method detail to Methods and keeping only the interpretive essentials.
### 5. Write captions for special figure types
1. **Micrographs/images:** scale bar (with length), imaging modality and settings, what the arrows/arrowheads mark, sample identity and n.
2. **Schematics/models:** define every symbol, arrow type, and color; state what the diagram asserts vs illustrates.
3. **Multi-panel composites:** describe panels in order with (a), (b), (c) labels matching the figure; give the shared methods once, panel-specific details per panel.
### 6. Quick-reference checklist
- [ ] Caption states what is shown AND what it means
- [ ] Panels described in order with matching (a), (b), (c) labels
- [ ] Sample sizes, statistics, error-bar definitions included
- [ ] Scale bars, arrows, colors, and symbols defined
- [ ] Every abbreviation defined at first use in the caption
- [ ] Consistent terminology with the main text
- [ ] Caption matches the final figure version (panel letters, counts)
- [ ] Length calibrated to the figure's information density
## Common pitfalls
- **Title-only captions:** "Effect of X on Y." — describes nothing; the reader learns zero.
- **Undefined symbols:** colors, stars, dashed vs solid lines unexplained — the most common caption failure.
- **Mystery error bars:** bars without "SD/SE/CI" labels are uninterpretable — always state.
- **Ambiguous n:** biological vs technical replicates unstated — inflates apparent rigor or hides pseudoreplication.
- **Caption–figure mismatch:** describing panels that don't exist, wrong panel letters, statistics for a different analysis — update captions last, after figures are final.
- **Interpretation overreach:** using the caption to make claims the figure doesn't support — the takeaway sentence must be earned by the panel.
- **Undefined abbreviations:** every abbreviation in the figure or caption defined at first use in the caption — captions must stand alone.
- **Caption–figure drift:** captions written for an earlier figure version (wrong panel letters, deleted panels described) — update captions last, after figures are final.