Visual Encoding

A visual encoding is the mapping from a number to a visual property: a position on an axis, the length of a bar, the angle of a slice, the area of a circle, the darkness of a shade. Every chart makes this choice, and the choice determines how accurately the number can be read back.

In 1984 William Cleveland and Robert McGill identified ten elementary perceptual tasks people use to extract quantitative information from graphs, and ordered them by accuracy. The ordering has six ranks: position along a common scale; positions along nonaligned scales; length, direction, angle; area; volume, curvature; and shading, color saturation.

Three of those ranks contain more than one task, and the authors are explicit about why: “at the moment there is not enough information to separate the ties.” The ordering is frequently quoted as a flat list of eight or ten individually ranked items, which asserts a precision the paper declines to claim.

Cleveland and McGill present the ordering as a hypothesis informed by their reasoning, experiments, and psychophysics. It guides the choice of encoding rather than prescribing a complete graph; apply it with judgment about the comparison the reader needs to make.

Used as guidance it is still the sharpest tool available, because it turns a preference into a decision rule: encode the comparison that matters using a task as high in the ordering as possible. The same three values read by position, by angle, and by volume are three different levels of reading accuracy — and the data is identical in all three.

References:

• Cleveland & McGill, Graphical Perception (JASA 79:387, 1984)