Shape is categorical only, and reliably distinguishable shapes are few

What it is

Shape-identity encoding generally uses different outlines to distinguish categories because circles, triangles, and crosses have no common high-to-low order. “Categorical only” refers to shape identity, not every geometric property: size, orientation, or aspect ratio may carry other relations. Reliable capacity has no task-independent number; it depends on the particular symbol set, size, density, legend, familiarity, medium, and required speed.

Why it happens

Shapes differ through contour, corners, notches, topology, and internal structure, and pairwise perceptual distances within a symbol set are usually uneven. Similar outlines become easier to confuse when small or crowded. Composite symbols also require readers to learn combination rules. Naming and domain conventions can improve recognition without making every added symbol equally separable.

Studying it

Test candidate symbols at the target size, background, and density, measuring category identification, pairwise confusion, search time, and legend lookup. Randomize category-to-shape mappings and separate first-use from trained performance. Use a confusion matrix to diagnose particular pairs; do not infer a universal capacity from one library or assume shape must support fewer or more categories than hue.

Where it stops holding

Experts may learn a large vocabulary in cartographic, chemical, or engineering symbol systems, but this depends on training, context, and stable convention. When order matters, use another ordered channel or a domain-semantic symbol sequence rather than inferring magnitude from arbitrary outline identity. Direct labels or facets may outperform an expanded symbol set.

Applying it

• Select structurally distinct candidates, then test confusion in the finished graphic rather than applying a fixed category limit.
• Use shape identity for category. If size or orientation has another meaning, separate those mappings in the legend and test interference.
• When categories or confusable pairs overwhelm the palette, aggregate, facet, label directly, or add filtering; retain nonvisual category names.
• Retest at the smallest screen, print, and zoom conditions, replacing symbols or reducing simultaneous categories where errors concentrate.

Related

• Same group: U1.07.2 Shape judgment needs adequate mark size; tiny scatter dots defeat it · U1.07.3 Texture fills are unstable under scaling and printing · U1.07.4 Orientation encodes a few ordered values but rotational symmetry confuses it · U1.07.5 Shape and texture differences survive greyscale output and colour-vision deficiency
• Nearby: U4.01 Categorical palette construction · U1.09.3 The categorical-colour limit is lower than the category count most charts actually use
• Search terms: shape encoding · visual channels · scatterplot symbols · nominal encoding