Every visual channel supports a limited number of reliably discriminable levels

What it is

A discriminable level limit is the number of levels that readers can reliably recover at a stated accuracy and speed for a specified channel, stimulus set, presentation condition, and task. It is not a constant attached to a channel name. Detecting a mark, deciding whether two marks differ, identifying an isolated mark, searching among many marks, and reading order or magnitude can yield different capacities. The finite resource is a reliably decodable relation under noise and a finite display—not the number of distinct values a data file can store.

Why it happens

As visual values occupy a channel more densely, perceptual response distributions for neighbouring stimuli are more likely to overlap. Display sampling, surround, crowding, memory, and the available comparison reference add uncertainty. A legend supplies an external reference but changes direct identification into repeated matching; it cannot remove confusion in the stimuli. Position, length, lightness, hue, and shape have different error structures, so a level count measured for one channel is not evidence for another.

Studying it

Fix the task and success criterion before varying level count or spacing. Detection asks whether a signal is present; discrimination asks whether a pair differs; identification uses a confusion matrix of presented and reported identities; search separately measures accuracy, time, and misses. Control or manipulate mark size, density, background, device, viewing distance, training, legend, and visual ability. Report psychometric curves and uncertainty intervals rather than only one maximum count.

Where it stops holding

A continuous axis may map many source values without making every value precisely distinguishable without labels. Ticks, text, and interactive lookup transfer some reading to other channels. Training and familiar domain symbols can improve a particular identification task without raising capacity for unfamiliar symbols. Anomaly detection and coarse trend judgment also require different resolution from naming every level.

Applying it

• Record the channel, task, simultaneous levels, smallest mark, background, device, and acceptance metric for each encoding; do not maintain a context-free table of channel ceilings.
• Validate detection, pair discrimination, isolated identification, and in-chart search separately, using the task the product actually requires to set levels.
• Report the worst stimulus pairs, error intervals, and participant variation. When evidence is weak, reduce simultaneous levels and add labels, filters, or structured values.
• Preserve source value, unit, category name, and order for nonvisual access so visual capacity does not become a data-access ceiling.

Related

• Same group: U1.09.2 Beyond the limit, added levels bring confusion, not information · U1.09.3 The categorical-colour limit is lower than the category count most charts actually use · U1.09.4 Discriminable-level limits drop as marks shrink · U1.09.5 Over the limit, aggregate categories instead of subdividing the encoding
• Adjacent: U1.08.1 Nominal, ordinal, and quantitative data need different encoding channels · U1.10.1 Encoding one variable on two channels improves judgment robustness
• Search terms: discriminable level limit · psychometric function · confusion matrix · visual channel capacity