Two different variables sharing one visual dimension interfere with each other

What it is

Channel collision and dimensional interference arise when two independently queried variables compete for one visual degree of freedom, or occupy separate but hard-to-attend integral dimensions. If both risk and amount determine lightness, one rendered value cannot reveal both sources. If amount controls lightness and probability controls chroma, nominally distinct mappings may still be read as overall “strength.” Neither case is redundant encoding, which maps the same variable twice.

Why it happens

A mark has only one realized size, lightness, or position at a time. Two inputs must be overwritten, prioritized, or combined into that value, so the mapping ceases to be invertible. Even with two channels, selective attention depends on whether the pair is separable, integral, or produces an emergent configural property. Garner interference describes irrelevant variation slowing or disrupting classification on the target dimension; it does not imply that every bivariate display fails.

Studying it

Use a full factorial design that independently varies A and B, asking participants to retrieve A, retrieve B, and judge their joint relation. Compare accuracy, response time, and systematic bias when the irrelevant dimension is fixed versus varying, and swap the target to detect asymmetric interference; record legend use. For a composite on one channel, also test whether different A/B pairs render identically and whether each source value can be recovered rather than merely ranking overall magnitude.

Where it stops holding

If the intended task is a predefined composite index, its combined value may legitimately occupy one channel, but the components can no longer be claimed as independently readable. Relatively separable pairs such as hue and shape still depend on size, category count, and task order. Correlated variables may create a useful visual pattern while masking poor component lookup; noticing correlation is not evidence of channel separability.

Applying it

• Maintain a variable–channel–task table marking which variables require independent lookup. When one channel takes several inputs, disclose the composition rule or separate the views.
• Prefer task-tested separable pairs for bivariate views and give each its own legend; use facets, juxtaposed charts, or switchable layers when needed.
• Do not assume that lightness with chroma or size with density is orthogonal. Ask participants to report each value while the other changes.
• Expose both source variables, units, and composition through keyboard and screen-reader access rather than announcing only a final color or “strength.”

Related

• Same group: U1.10.1 Encoding one variable on two channels improves judgment robustness · U1.10.2 Redundant encoding is the main fallback for colour-vision deficiency and greyscale output · U1.10.4 Redundant encoding spends channels that other variables could use · U1.10.5 Redundant channels that disagree semantically give contradictory cues
• Adjacent: U1.06.4 Lightness and saturation changes are read as one strength dimension · U1.08.1 Nominal, ordinal, and quantitative data need different encoding channels
• Search terms: channel collision · Garner interference · integral dimensions