A scientific figure is any visual element in a research paper that presents data, illustrates a concept, or shows an image. Figures are often the first thing readers look at when scanning a paper, and many people review the figures without reading the full text. They serve as the visual backbone of scientific communication, translating complex findings into something a reader can absorb at a glance.
Types of Scientific Figures
Scientific figures fall into two broad categories: statistical and non-statistical. Statistical figures are what most people picture first. These include line graphs, bar charts, scatter plots, pie charts, and other data visualizations that display numerical results. A line graph might track how a measurement changed over time, while a bar chart compares values across different groups.
Non-statistical figures cover everything else. Clinical images and photographs, such as X-rays, CT scans, MRI images, microscopy findings, and photographs of tissue samples, provide direct visual evidence from individual cases. Diagrams like flowcharts, algorithms, pedigree charts, and maps display complex relationships or processes. Some figures are purely illustrative, showing the structure of a molecule, the layout of an experimental setup, or a conceptual model of how a biological system works.
What Makes Up a Figure
A well-built scientific figure has several standard components that help the reader understand it without hunting through the text.
- Panels: Many figures contain multiple related images or graphs arranged together. Each panel is labeled with a capital letter (A, B, C) so the text can refer to specific parts. When panels need further subdivision, the convention is to use Roman numerals next (Ai, Aii, Bi, Bii) rather than repeating lowercase letters.
- Axis labels and units: Any graph includes clearly labeled axes with appropriate units (hours, micrometers, concentration). Without these, the data is meaningless.
- Scale bars: Microscopy images and photographs include a small bar indicating physical size, since magnification alone doesn’t tell you how large something actually is.
- Legends and annotations: Labels placed directly on the figure itself, such as arrows pointing to structures or text identifying experimental conditions, reduce how much the reader needs to flip between the image and the caption. The goal is for the figure to be understandable on its own.
Below the figure sits the caption, sometimes called the figure legend. A good caption starts with a single-sentence title that states the figure’s main message. The rest describes, at a high level, what was done to generate the data. Captions stick to description rather than interpretation, since the discussion of what the results mean belongs in the paper’s text.
When to Use a Figure Instead of a Table
Figures and tables both present information, but they serve different purposes. Tables work best when readers need to look up exact values or compare precise numbers across many categories. Figures are better when the point is a trend, a pattern, a comparison of shapes, or a visual relationship that would be hard to describe in rows and columns. If you’re showing that something increased over time, a line graph communicates that instantly. If you’re showing what a tumor looks like under a microscope, no table can replace the image. Flowcharts and diagrams also belong as figures because they map out processes or relationships that are spatial by nature.
Resolution and Image Quality
Journals have strict technical requirements for figure quality. Most require a minimum resolution of 300 pixels per inch (ppi). Figures with lots of small text, detailed charts, or multiple photographs render best at 600 ppi. Anything below 300 produces blurry, jagged, or pixelated images in the published paper.
One common mistake is placing a low-resolution element inside a high-resolution file. If you create a graph at 72 ppi (typical screen resolution) and drop it into a 300 ppi image file, the graph will still look blurry. The overall quality of a figure is only as good as its lowest-resolution component.
Color and Accessibility
Roughly 8% of men and 0.5% of women have some form of color vision deficiency, which means color choices in figures matter more than most people realize. The most important rule is to avoid pairing red and green in the same image, since that’s the combination most commonly confused by people with color blindness.
Better two-color alternatives include green and magenta, yellow and blue, or red and cyan. For figures with three or four colors, magenta/yellow/cyan or magenta/yellow/green/blue work well. Heatmaps that show a range of values should use two complementary colors at the extremes with white or black in the middle. Scales like green-to-black-to-magenta or blue-to-white-to-red are easier for color-blind readers than rainbow spectrums.
Beyond color choice, adding different shapes, line styles, or symbols to represent different data groups ensures the information comes through even in grayscale. When possible, presenting data in black, white, or grayscale is the simplest solution. Color should add meaning, not just decoration.
Ethics of Image Editing
Scientific figures follow strict ethical guidelines about what kinds of editing are acceptable. The core principle, as stated by the Journal of Cell Biology, is that adjustments must not “obscure, eliminate, or misrepresent any information present in the original.”
Reasonable adjustments include minor changes to brightness and contrast applied uniformly across the entire image. Brightness shifts all the pixel intensities lighter or darker, while contrast expands or compresses the range between the lightest and darkest areas. Large adjustments to either one risk clipping information at the extremes, making bright areas pure white or dark areas pure black, which permanently removes detail. Auto-correction tools in image editing software are discouraged because they tend to over-process images in ways that are difficult to document or reproduce.
Cropping is generally acceptable for centering a subject, trimming empty space, or removing a stray piece of debris. It crosses an ethical line when it changes the context of what remains. Cropping out dead cells to show only healthy ones, or removing bands from a gel image that contradict a hypothesis, constitutes data manipulation. The distinction is straightforward: removing irrelevant background is fine, removing inconvenient data is not.
Tools for Creating Figures
The software researchers use depends on the type of figure. For data-driven graphs and plots, programming tools like R and Python (with visualization libraries) are standard in many fields because they produce reproducible, publication-quality graphics. For biological diagrams and conceptual illustrations, dedicated platforms like BioRender have become widely adopted, with figures appearing in over 300,000 citations across 1,500 institutions. General-purpose tools like Adobe Illustrator handle vector graphics for custom diagrams and figure assembly, while ImageJ is a go-to for processing microscopy images. Many researchers still start in presentation software for quick drafts, but the final versions typically need dedicated tools to meet journal specifications for resolution, file format, and labeling.