Spatial distance between related elements is cognitive distance
Aliases: proximity compatibility · grouping by nearness · cognitive distance
What it is
A price sitting against a product name is read as that product’s price. The same digits parked in a footer “subtotal” have to be claimed before they can enter a comparison. Gestalt proximity treats spatial distance as the first evidence of “same thing”: near means related, far means someone else’s.
Label and field, legend and the line it names, avatar and username, are compiled into one object by that distance. Split them and the layout is not merely uglier — working memory is holding an unclaimed fragment.
Why it happens
Visual cortex is extremely sensitive to distance; grouping finishes before word meaning is identified. Proximity is the cheap cue, arriving ahead of borders, fills, and headings. Wickens’ proximity compatibility states it as a processing rule: sources that must be used together should sit together; sources that must be processed separately should be pulled apart. Distance is then glue or a buffer, not leftover aesthetic margin.
An unrelated element inserted on the path steals the glue — an error message with a recommendation wedged between it and its field will be read as part of the recommendation.
Studying it
Use a grouping judgement: the same elements under two or three spacing schemes, “which of these are one thing?”, or a task that forces the two ends to be used together (fill from a label, read a value from a legend).
Independent variables: ratio of within-group to between-group spacing, presence of an intruder, presence of compensating cues such as a border. Dependent variables: grouping accuracy, how often an unrelated item is absorbed, time on the integration task.
Eye tracking can show whether the two ends form one fixation cluster; mis-claim errors in behaviour are closer to the real cost.
Where it stops holding
Two prices that must be compared, if packed too close, fuse into “one price” and the separation the comparison needed is eaten by proximity. Tables work by grid alignment; nearness between cells must not be read as “same field”. In a reversed reading direction the default side of “near” flips, and a left-stuck label sends the relation to the wrong side. Screen readers group by source order, not pixel distance; visual proximity does not travel, and programmatic labelling has to carry it.
Applying it
- List the pairs that must be understood together — label–field, name–price, legend–series — and make each pair’s gap clearly smaller than the gap to the next pair.
- Stick the error copy to the field that failed, not to a dump at the foot of the form.
- Pull items that must be compared far enough that they will not be read as one object, then align them.
- How to check: scramble the wording, keep the positions, and ask people to circle “what belongs together”. Every wrong circle is distance lying.