Captions That Work
Definition
A narrative caption is a multi-sentence, stand-alone caption that turns a figure into a small, self-contained story: what’s being shown, how it was produced, and what it means. This is distinct from the short, purely labeling caption most people default to (“Figure 1. Growth rates of Strain A and Strain B at different temperatures”), which tells a reader what they’re looking at but not why it matters.
This isn’t an idiosyncratic preference — it’s grounded in how style guides and working scientists actually write. Style guides call it the “extended” or “narrative” caption. Edward Tufte argues for figures with rich, integrated explanatory text rather than bare labels. The CSE Manual’s guidance for scientific writing is explicit: a figure legend should be complete enough that a reader can understand the figure without reading the surrounding text. Joshua Schimel frames it as “telling the story of the figure.” Nature and Science both expect multi-sentence legends that explain what’s shown, how it was generated, and what it means.
Learning Outcome
After using this technique, you should be able to write a caption that stands on its own — giving context, describing the method or metric behind the data, and interpreting the result — and to scale that caption’s length and tone appropriately for the venue you’re writing for.
Core Structure
A strong stand-alone caption does four things, in this order:
- Start with context — what is being shown, and why does it matter?
- Describe the method or metric — how was the data obtained or calculated?
- Interpret the result — what should the reader take away from it?
- Reference the main text only if needed — for readers who want more depth, not as a crutch for the caption itself.
This matters because of how people actually read technical papers: they skim by jumping from figure to figure before committing to the full text. A caption that only labels forces the reader back into the prose to understand what they’re looking at. A narrative caption lets the figure carry its own weight — which also makes it easier to reuse in talks, reports, or teaching material later, independent of the paper it came from.
Scaling by audience. The same figure gets a differently-shaped caption depending on where it’s going:
| Audience | Focus | Length | Tone |
|---|---|---|---|
| Journal article | Accuracy, replicability | Medium (~4–6 sentences) | Formal, technical |
| Technical report | Completeness, documentation | Long (~5–10 sentences) | Formal, readable |
| Grant proposal | Significance, impact | Medium (~3–5 sentences) | Narrative, persuasive |
| Conference presentation | Fast comprehension | Short (~2–3 sentences) | Crisp, conversational |
Worked Example
Same figure — bacterial growth rates for two strains across a temperature range — written four different ways.
Short (labeling) caption, for comparison:
Figure 1. Growth rates of Strain A and Strain B at different temperatures.
Journal article (medium, formal/technical):
Figure 1. Growth rates of Strain A and Strain B across temperatures from 10°C to 40°C, measured in triplicate under controlled laboratory conditions. Strain A peaks near 30°C and declines sharply above it, indicating a narrow thermal tolerance window; Strain B holds a stable growth rate across the full range, consistent with broad thermal adaptation. Error bars represent standard deviation. These results may inform strain selection for thermally variable industrial applications.
Grant proposal (medium, narrative/persuasive):
Figure 1. Strain A thrives at 30°C but declines sharply thereafter, while Strain B maintains steady growth across the full temperature range tested. This robustness suggests Strain B could support reliable bio-production in environments with variable temperature.
Conference presentation (short, crisp):
Figure 1. Strain A grows best at 30°C but drops off sharply at higher temperatures. Strain B grows steadily across a wide range — the kind of thermal tolerance that matters for variable-temperature bio-industrial use.
Notice what’s constant across all four: the core claim (Strain B is more thermally resilient) and the interpretation (why that matters). What changes is length and how directly the significance is stated. The short label version, by contrast, requires the reader to go find that story somewhere else in the paper.
Common Pitfalls
- Restating what’s already obvious from the figure or the main text — a caption should add context and interpretation, not just repeat the axis labels in words.
- Describing the figure but never interpreting it — “the line goes up” is not the same as “the line going up means X.”
- Using the same caption style regardless of venue — a conference-slide caption dropped into a journal submission reads as thin; a journal-length caption on a slide will lose the room.
- Leaving the caption unconnected to the actual research question the figure is meant to support.
Rubric / Checklist
A good figure legend:
- Gives sufficient background without being redundant with the main text
- Focuses the reader’s attention on the important pattern or anomaly, not just “what’s plotted”
- States or clearly implies an interpretation, not just a description
- Relates explicitly back to the research question or hypothesis
- Is scaled in length and tone to its actual venue (see the audience table above)