The Whys - Figure Checklist

The whys - the science of what works for chart figures

My article on chart figures is out in Nature Cell Biology. But, the preprint was of course longer. I had included the data and research we have on why each step of the checklist is important. That was cut from the final article, but I feel it is still nice to know for some of you. Therefore, here are the rationales, the reasons for what we do, the whys.

BASICS

Choose chart

How charts work is an active area of research in computational sciences. For example, this revealed that humans excel at interpreting charts with a common scale, such as bar (x-axis) and scatter plots (x- and y-axis), but this task is harder with charts lacking a shared axis, like line and pie charts (Cleveland and McGill, 1984). And the task vary with chart type: line charts require comparison of direction and angle, maps and heat maps of color or texture, and treemaps and bubble charts of area (Cleveland and McGill, 1984); the success of the latter furthermore depends on the choose shape, with circles being harder to interpret than rectangles (Kong et al., 2010). Comparing trends across multiple pie charts is less effective than using a single bar or line chart (Hollands and Spence, 1992). Recent studies highlight challenges with bar charts that summarize distributions, as they obscure data transparency (Weissgerber et al., 2015). A great review of data visualization science is Franconeri et al., 2021.

• Cleveland, W.S., McGill, R., 1984. Graphical Perception: Theory, Experimentation, and Application to the Development of Graphical Methods. Journal of the American Statistical Association 79, 531–554. here
• Franconeri, S.L., Padilla, L.M., Shah, P., Zacks, J.M., Hullman, J., 2021. The science of visual data communication: What works. Psychological Science in the public interest 22, 110–161.
• Hollands, J.G., Spence, I., 1992. Judgments of Change and Proportion in Graphical Perception. Hum Factors 34, 313–334. here
• Kong, N., Heer, J., Agrawala, M., 2010. Perceptual Guidelines for Creating Rectangular Treemaps. IEEE Transactions on Visualization and Computer Graphics 16, 990–998. here
• Weissgerber, T.L., Milic, N.M., Winham, S.J., Garovic, V.D., 2015. Beyond bar and line graphs: time for a new data presentation paradigm. PLoS Biol 13, e1002128. here

Simplify charts

How much data is too much highly depends on the audience, thus generalizing studies are rare. Research however did show that increasingly dense versions of time course data remained legible even at reduced sizes (Heer et al., 2009). A recently widely shared chart, the climate stripes, even condenses a trend into merely colored stripes, retaining its information content (Hawkins, n.d.). Simplifying visual information, is a key recommendation from Edward Tufte, who introduced this strategy as “small multiples” (Tufte, 2011). Small multiples can be easily realized in programming software used for data visualization such as Python and R, where the feature is available as “gridding” and “faceting” respectively.

• Hawkins, E., n.d. Show Your Stripes [WWW Document]. Show Your Stripes. here
• Heer, J., Kong, N., Agrawala, M., 2009. Sizing the horizon: the effects of chart size and layering on the graphical perception of time series visualizations. Proceedings of the SIGCHI Conference on Human Factors in Computing Systems 1303–1312. here
• Tufte, E.R., 2011. The Visual Display of Quantitative Information. Graphics Press, Cheshire, Connecticut.

Text

Figures are most memorable when they integrate text with familiar keywords and phrases (Hegarty and Just, 1993; Martin, 2020). Comprehension is key, a study found that concise, jargon-free journal titles with minimal abbreviations correlate with higher citations (Letchford et al., 2015). Clarity/memorability can also be enhanced by supplementing text with icons or pictograms (Haroz et al., 2015). Typography is crucial for legibility, though scientists are often limited publishers and institutional formatting guidelines. When choosing a font, ensure it is legible, with clear ascenders, descenders, and adequate white space (Wong, 2011a). A legible font features consistent letter height (x-height), clear ascenders (e.g., in “b”) and descenders (e.g., in “p”), and sufficient white space within closed letters (e.g., “a” and “o”). There is no clear consensus on whether serif (e.g., Times New Roman) or sans serif (e.g., Arial) fonts are more readable (Arditi and Cho, 2005).

• Arditi, A., Cho, J., 2005. Serifs and font legibility. Vision Research 45, 2926–2933.here
• Haroz, S., Kosara, R., Franconeri, S.L., 2015. ISOTYPE Visualization: Working Memory, Performance, and Engagement with Pictographs, in: Proceedings of the 33rd Annual ACM Conference on Human Factors in Computing Systems, CHI ’15. Association for Computing Machinery, New York, NY, USA, pp. 1191–1200. here
• Hegarty, M., Canham, M.S., Fabrikant, S.I., 2010. Thinking about the weather: How display salience and knowledge affect performance in a graphic inference task. Journal of Experimental Psychology: Learning, Memory, and Cognition 36, 37–53. here
• Hegarty, M., Just, M.A., 1993. Constructing mental models of machines from text and diagrams. Journal of Memory and Language 32, 717–742. here
• Letchford, A., Moat, H.S., Preis, T., 2015. The advantage of short paper titles. R Soc Open Sci 2, 150266. here
• Martin, K., 2020. A picture is worth a thousand words. MEW 29, 28–34.
• Wong, B., 2011a. Points of View: Typography. Nat Methods 8, 277. here

DESIGN

Layout

It is important to think how to arrange elements on a poster and a in multi-panel figures – this ensures readers see most important elements first. Similarly, arranging information efficiently in a chart is critical – a legend should be close to the item it explains, a a title should be findable. Further resources on how to organizing figure panels (Wong, 2011b, 2011c), graphical abstracts (Jambor and Bornhäuser, 2024) and posters (Brown, 1996; Faulkes, 2021; Jambor, 2023) is available, as well as tools that help authors quickly arranging items in a figure panel (e.g., in ImageJ: Aigouy and Mirouse, 2013; Mutterer and Zinck, 2013).

• Aigouy, B., Mirouse, V., 2013. ScientiFig: a tool to build publication-ready scientific figures. Nat Methods 10, 1048.here
• Brown, B.S., 1996. Communicate your science! … Producing punchy posters. Trends Cell Biol 6, 37–39. here
• Faulkes, Z., 2021. Better Posters: Plan, Design and Present an Academic Poster. Pelagic Publishing Ltd.
• Jambor, H.K., 2023. Insights on poster preparation practices in life sciences. Front Bioinform 3, 1216139. here
• Jambor, H.K., Bornhäuser, M., 2024. Ten simple rules for designing graphical abstracts. PLOS Computational Biology 20, e1011789. here
• Mutterer, J., Zinck, E., 2013. Quick-and-clean article figures with FigureJ. Journal of Microscopy 252, 89–91. here
• Wong, B., 2011b. Layout. Nat Methods 8, 783. here
• Wong, B., 2011c. Points of view: Points of review (part 1). Nature Methods 8, 101–101. here

Color schemes, Encode data with color

“Color” is defined by three parameters: hue (red, green, blue etc), saturation (bright red, pale red), and lightness (dark-light) (Wong, 2010a). As a core preattentive attribute, color has an immediate impact on perception. When using color to represent quantities, it is crucial to choose a palette that accurately reflects the data (Gehlenborg and Wong, 2012; Hattab et al., 2020; Szafir, 2018). This is especially important when sequential colors are encoded with multi-color palettes. Oftentimes, jet or rainbow color palettes were used for this case, however these vary both the color as well as the lightness – the result are inhomogeneous perceived color changes (border effects), that do not faithfully represent the sequential chances in the data. Addressing this issue, new multi-hued palettes were developed, such as Viridis, which do not vary in lightness (Smith and van der Walt, n.d.). A helpful tool for selecting data-encoding colors is Cynthia Brewer’s work, originally created for cartography but now widely used in various scientific fields (Brewer, n.d.; Harrower and Brewer, 2003).

When selecting color schemes consider tools like Paletton (https://paletton.com/) to pick harmonious or contrasting color combinations that enhance the visual appeal of your work. Tools and simulations are essential for verifying your colors work for all audiences. Note that color perception varies even among visually able audiences. Most computers have settings to simulate color blindness and greyscale, alternatively web-based tools are also available. Use WebAIM’s contrast checker to ensure a minimum contrast ratio of 4.5 to 1. Refer to previous work for more details on color accessibility for audiences and tasks (Hattab et al., 2020; Kim et al., 2021) and select harmonious or contrasting color schemes with tools like Paletton (https://paletton.com/).

• Brewer, C.A., n.d. ColorBrewer: Color Advice for Maps. here
• Gehlenborg, N., Wong, B., 2012. Mapping quantitative data to color. Nature Methods 9, 769–769. here
• Harrower, M., Brewer, C.A., 2003. ColorBrewer.org: An Online Tool for Selecting Colour Schemes for Maps. The Cartographic Journal 40, 27–37. here
• Hattab, G., Rhyne, T.-M., Heider, D., 2020. Ten simple rules to colorize biological data visualization. PLOS Computational Biology 16, e1008259. here
• Kim, N.W., Joyner, S.C., Riegelhuth, A., Kim, Y., 2021. Accessible Visualization: Design Space, Opportunities, and Challenges. Computer Graphics Forum 40, 173–188. here
• Smith, N. & Van der Walt, S., 2015. A Better Default Colormap for Matplotlib. here
• Szafir, D.A., 2018. Modeling Color Difference for Visualization Design. IEEE Transactions on Visualization and Computer Graphics 24, 392–401. here
• Wong, B., 2010a. Color coding. Nature Methods 7, 573–573. here

On beauty

There is no formal framework for “beauty” in data visualization, nor research whether prettier charts are more insightful or memorable. However, the Gestalt principles help explain how we interpret patterns (Wertheimer, 1912). We naturally group objects based on similarity, proximity, or connection by lines. Symmetry also plays a key role, as symmetrical elements are perceived as belonging together. This principle is often applied in the organization of panel figures. In bar charts, ordering bars by quantity enhances both clarity and aesthetic appeal. By incorporating symmetry, we not only improve the effectiveness of a visualization but also make it more visually pleasing. While beauty in visualization is subjective, the role of emotion is well-studied. Research shows that positive emotions elicited by a chart can boost user engagement (Kennedy and Hill, 2018) and data animations, designed to evoke emotions like joy or amusement, increase attention and interest (Lan et al., 2022). Humans are also drawn to familiar visuals: charts incorporating images or pictograms—sometimes dismissed as “chart junk”—were found to be more re-recognizable in a large recall study (Borkin et al., 2013). And complementing text with relevant images can enhance memorability (Haroz et al., 2015) and likely increase user engagement.

• Borkin, M.A., Vo, A.A., Bylinskii, Z., Isola, P., Sunkavalli, S., Oliva, A., Pfister, H., 2013. What makes a visualization memorable? IEEE Trans Vis Comput Graph 19, 2306–2315.here
• Kennedy, H., Hill, R.L., 2018. The Feeling of Numbers: Emotions in Everyday Engagements with Data and Their Visualisation. Sociology 52, 830–848. here
• Lan, X., Shi, Y., Wu, Y., Jiao, X., Cao, N., 2022. Kineticharts: Augmenting Affective Expressiveness of Charts in Data Stories with Animation Design. IEEE Transactions on Visualization and Computer Graphics 28, 933–943. here
• Wertheimer, M., 1912. Experimentelle Studien uber das Sehen von Bewegung. Zeitschrift fur Psychologie 61, 161–165.

FEEDBACK

The 1-second test: “What do you see first?”

Human visual perception is limited. This was first described by Miller (1956) as the “Magic Number 7,” highlighting our ability to perceive a maximum of 7 visual elements at a time (Miller, 1956). Today, this is referred to as “chunking” in learning, and from human vision research we know that our eyes are biased towards certain features, so called preattentive attributes like color, size, or shape, that attract attention more easily (Wong, 2011d, 2010b). In figures, objects with most salience thus should be the most relevant (Canham and Hegarty, 2010; Hegarty et al., 2010).

• Canham, M., Hegarty, M., 2010. Effects of knowledge and display design on comprehension of complex graphics. Learning and Instruction, Eye tracking as a tool to study and enhance multimedia learning 20, 155–166. here
• Miller, G.A., 1956. The magical number seven, plus or minus two: Some limits on our capacity for processing information. Psychological Review 63, 81–97. here
• Wong, B., 2011d. Salience to relevance. Nat Methods 8, 889. here
• Wong, B., 2010b. Salience. Nat Methods 7, 773. here

Reverse feedback. Ask: “Explain to me what you see.”

Reverse feedback is similar to expert feedback, a method established in computer interface design (Shneiderman and Plaisant, 2010). In the laboratory setting, eye tracking may be used, while A/B testing compares different versions. Expert feedback allows authors to understand usability and potential issues (Tory and Moller, 2005) and is similar to heuristic evaluations, where usability principles are used to standardize evaluations (Zuk et al., 2006), similar to the figure checklist (Figure 1).

• Shneiderman, B., Plaisant, C., 2010. Designing the user interface: strategies for effective human-computer interaction. Pearson Education India.
• Tory, M., Moller, T., 2005. Evaluating visualizations: do expert reviews work? IEEE Computer Graphics and Applications 25, 8–11. here
• Zuk, T., Schlesier, L., Neumann, P., Hancock, M.S., Carpendale, S., 2006. Heuristics for information visualization evaluation, in: Proceedings of the 2006 AVI Workshop on BEyond Time and Errors: Novel Evaluation Methods for Information Visualization, BELIV ’06. Association for Computing Machinery, New York, NY, USA, pp. 1–6.here

Focus the attention

Visual perception research, starting with Bauhaus defined Gestalt principles (Wertheimer, 1912) and evolving through Miller’s “Magic Number 7”, shows that we process visual information in chunks (Johnson, 1970; Miller, 1956). Bertin expanded this and defined the simplest elements we see as the “pre-attentive attributes” (Bertin, 1967). Hubel and Wiesel studied the neurobiology of the visual system, which confirmed that basic cues such as color, orientation, size or motion are processed by neurons for in the visual cortex, while complex juxtapositions (e.g. charts) are integrated in a later stage of the brain (Barlow, 1982). Focusing on salient elements in figures, as Tufte suggests, aligns with how the brain processes visual data (Tufte, 2011).

• Barlow, H.B., 1982. David Hubel and Torsten Wiesel: Their contributions towards understanding the primary visual cortex. Trends in Neurosciences 5, 145–152. here
• Bertin, J., 1967. Sémiologie graphique | Le Comptoir Des Presses d’Universités.
• Johnson, N.F., 1970. The Role of Chunking and Organization in The Process of Recall1, in: Bower, G.H. (Ed.), Psychology of Learning and Motivation. Academic Press, pp. 171–247. here
• Miller, G.A., 1956. The magical number seven, plus or minus two: Some limits on our capacity for processing information. Psychological Review 63, 81–97. here

Download the checklists here