Figure Design

Criteria for success

1. Your figure leaves the audience with a clear, one-sentence main message
2. You provide evidence that directly supports the main message
3. Any content distracting from your main message is removed from the figure
4. Visual components are easily distinguished

Structure Diagram

Purpose

Figures are any visual presentations of results and come in a variety of forms. Graphs, diagrams, photos, drawings, schematics, and maps are all types of figures.  Despite this variety, the purpose of all figures is to communicate two things:

1. A take-away message
2. Evidence showing your message is true

Your message is the reason you want to show people your data! The purpose of a figure is not only to show people your data: it should also communicate the message you have learned by interpreting that data.

Analyze your audience

An accurate analysis of your audience enables you to convincingly deliver your message. Ask yourself who will be viewing this figure and whether they are specialized in your field or represent a more general audience. Depending on the answer, you will want to tailor the complexity of your message, the conventions and language that you use, and the amount of evidence you present.

For example, if your figure is for publication in a field-specific journal, your audience will likely have significant background knowledge about your topic. In this case, you can deliver a very specific message (e.g., “at all bulk Reynolds numbers, the addition of woven meshes in the flow-channels increases the mean Nusselt number”) and provide detailed evidence (e.g., Nusselt number vs Reynolds number plot, with and without mesh). Insufficient or excessively simplified evidence will result in skepticism about your message.

If your figure will be used in a presentation for a high school outreach program, your audience will have a very different set of background knowledge. Here, you should alter the message (e.g., “woven meshes in channels increase heat loss”) to make it more interpretable and can present much less evidence (e.g., inlet-outlet temperature difference, with and without mesh). Giving a general audience too much evidence will be overwhelming; they won’t be able to distinguish the evidence that supports your main message from all the other details.

Skills

Choose figure designs that best communicate your message

Just as different words may be better or worse at communicating an idea, different figure designs may be better or worse at communicating your message. In designing a successful figure, consider which media, figure types, and plot types (see below for examples of each) best highlight your message.

For complex messages, multiple panels can break down a message into clear statements. Multi-panel figures will likely employ a combination of media and plot types. Use the complementary strengths of each element to communicate your message. Consider the following example.

In electrodialysis, an electric field is applied across repeating cell-pair units to separate a saline feed into diluted and concentrated streams (left). The cell-pairs form an electrical circuit, with the transport of ions represented by the current density i (right).

Media can deliver identical messages, but do so differently.

• Text and speech deliver precise statements.
• Tables list information with little context or interpretation.
• Figures illustrate conclusions with evidence and are open to interpretation.

As an example, compare the table and the graph below to the statement, “Increasing the time-averaged current density ratio decreased total specific energy consumption, because the decrease in pumping energy consumption exceeded the increase in desalination energy consumption.” While both contain the exact same data, the figure suggests an interpretation and makes it easier to convey trends in the data.

Table 1: Specific energy consumption at varying time-averaged current density ratios

Figure types present different forms of information.

• Photos portray the subject exactly, providing concrete evidence.
• Illustrations relax precision to draw attention to a chosen theme or element.
• Graphs display processes, quantities, or comparisons.

(Figures courtesy of Jungki Song, Space Nanotechnology Lab, MIT)

Plot types emphasize different types of data.

• What are you trying to show with your data: a correlation, a distribution, an event in time?

When you have a distribution on hand, using a summary of your data (i.e. mean and standard deviation) can obscure interesting information about your data. For example, the four datasets below have the exact same mean and similar standard deviations. This is Anscombe’s quartet, a group of datasets which are used to demonstrate the importance of graphing individual data points before deriving conclusions from descriptive statistics, such as the mean and standard deviation.

Source: adapted from the Anscombe’s Quartet illustration on Wikipedia.

Tweak your figure depending on the setting

WIll you be presenting your figure in an academic paper, a poster presentation, an oral presentation…? The final format dictates how your audience will interact with the figure, and how much support or explanation you will be able to provide.

Maximize your signal-to-noise ratio

Treat the message you want to communicate as your “signal”. Your goal is to transmit this signal as clearly as possible to your audience. Anything that interferes with communication of your message is “noise.” We’ve discussed ways to increase your signal by optimizing figure design. Here we will talk about strategies for minimizing noise.  

Noise from evidence

Don’t drown your audience in data: include only the minimum data necessary to make your point. Including evidence that doesn’t directly support your message distracts from evidence that does. Note that error bars do not constitute noise, and should be included to indicate the uncertainty when reporting measured values.

Noise from presentation

The way you present your chosen evidence can also draw attention away from your message. Using the figure below, we’ll give a few common examples of how a figure can be improved to remove noise.

1. The title of the figure is changed from a description of the data to a message about the data.
2. Legends are moved directly next to the data they describe, so the reader doesn’t have to look back and forth and match colors.
3. Color scheme is simplified and changed to draw attention to the relevant portion of data.
4. Unnecessary 3D graphics are removed.

Source: Trees, maps, and theorems (Jean-Luc Doumont), page 99

Many other types of noise exist. For example, unnecessary gridlines or axis labels can clutter a figure. Ask yourself what you want your audience to take away from the figure, and how you can make it easier for them to locate and focus on the relevant information.

(The metaphor of “signal-to-noise ratio” comes from Jean-luc Doumont’s book Trees, Maps, and Theorems.)

Resources for advanced figure design

There are many great resources for styling visual communications and data. A small selection is listed below. Additional articles on figure design can be found in this collection of articles from Nature Methods.

General figure design:

• Ten Simple Rules for Better Figures, from PLOS
• What makes bad figures bad?, a useful summary of figure design rules
• Elements of Visual Style, from Nature Methods
• Advice on plot elements: Axes, ticks and grids, plotting symbols

Composition:

• Layout, from Nature Methods
• White space, from Nature Methods

Color:

• Color coding, form Nature Methods
• Mapping quantitative data to color, from Nature Methods

Coding:

• Blog with tutorials on the technical aspect of making figures, including examples in MATLAB, R, Illustrator

Sample code for better-looking MATLAB plots, adapt to your own purpose

Resources and Annotated Examples

Annotated Example 1

This figure illustrates effective use of color, layout, and wording to convey its message. 462 KB

Annotated Example 2

This example discusses strategies to make clear plots without extraneous noise. 386 KB