Shape It Up: An Empirically Grounded Approach for Designing Shape Palettes

TL;DR

A model based on pairwise relations between shapes in the authors' experiments and the number of shapes required for a given design is developed, which advances understanding of shape perception in visualization contexts and provides practical design guidelines that can help improve categorical data encodings.

Abstract

Shape is commonly used to distinguish between categories in multi-class scatterplots. However, existing guidelines for choosing effective shape palettes rely largely on intuition and do not consider how these needs may change as the number of categories increases. Although shapes can be a finite number compared to colors, they can not be represented by a numerical space, making it difficult to propose a general guideline for shape choices or shed light on the design heuristics of designer-crafted shape palettes. This paper presents a series of four experiments evaluating the efficiency of 39 shapes across three tasks -- relative mean judgment tasks, expert choices, and data correlation estimation. Given how complex and tangled results are, rather than relying on conventional features for modeling, we built a model and introduced a corresponding design tool that offers recommendations for shape encodings. The perceptual effectiveness of shapes significantly varies across specific pairs, and certain shapes may enhance perceptual efficiency and accuracy. However, how performance varies does not map well to classical features of shape such as angles, fill, or convex hull. We developed a model based on pairwise relations between shapes measured in our experiments and the number of shapes required to intelligently recommend shape palettes for a given design. This tool provides designers with agency over shape selection while incorporating empirical elements of perceptual performance captured in our study. Our model advances the understanding of shape perception in visualization contexts and provides practical design guidelines for advanced shape usage in visualization design that optimize perceptual efficiency.

Figures (9)

• Figure 1: (a) Two examples of stimuli used in Experiment 1, both with six categories encoded with single-type (unfilled) and two-type (filled + open). (b) Two scatterplots used in Experiment 2, encoded with different shape palettes from Matlab and Tableau, both with six categories. (c) Two scatterplots with different category numbers (3 and 6) used in Experiment 4 for measuring pairwise distances. Both (a) and (b) employed relative mean judgment tasks while (c) applied correlation judgment tasks.
• Figure 2: We collected shapes from multiple sources and categorized them into three shape types: filled, unfilled, and open. Both filled and unfilled have 10 shapes and open type has 7 shapes.
• Figure 3: The average accuracy of mean judgment task separated by different shape types and type group combinations in Experiment 1. Overall group means are indicated in red, with category number on the x-axis. We used the scale from 30%-100% as the slope of the graph (i.e., how robust each group is to increasing numbers of categories) is the key signal in the data. Error bars represent 95% confidence intervals. Note that open shapes only supported 2--7 categories.
• Figure 4: We selected five shape palettes from common visualization tools, including Tableau, Matlab, R, Excel, and D3. Tableau, Matlab, and R have 10 shapes, Excel has 9 shapes and D3 has 7 shapes.
• Figure 5: The average mean judgment task accuracy of different shape palettes (top) from professional tools in Experiment 2, with category number on the x-axis. Each chart presents the average accuracy broken down by category numbers (aggregate accuracy is presented in red). The accuracy drops as the category number increases. The average accuracy ranges from 78% to 92%. Error bars represent 95% confidence intervals. Note that Excel (d) and D3 (e) have 9 and 7 shapes respectively.