Non-data elements consume attention while carrying no information

What it is

A task-irrelevant non-data element neither varies with the data nor helps readers interpret scale, unit, structure, provenance, uncertainty, annotation, or interaction, yet may consume space and attention. Tufte's data-ink ratio and chartjunk are heuristics for auditing graphical redundancy, not empirical laws that every reduction in non-data pixels improves a chart. “Carrying no information” applies only to elements with no demonstrable function in the stated task—not to every border, tint, word, or illustration.

Why it happens

Salient but irrelevant lines, shadows, textures, or colour differences may compete with marks, imply false groups, depth, or magnitude, and add legend matching or search. Cost depends on similarity to the target, salience, density, and task; decoration does not capture attention on every occasion. Structural cues, narrative imagery, and semantic prompts may instead improve localization, memory, or engagement. Failing to encode a number is therefore not sufficient evidence that an element conveys nothing.

Studying it

Create a baseline and controlled add/remove variants of the same data while preserving scale, layout, labels, and content. Test value reading, comparison, anomaly search, conclusion recall, and trust calibration; measure accuracy, response time, misses, gaze allocation, and reported effort. If embellishment is under study, separate immediate comprehension, delayed memory, and engagement. Experiments by Skau and colleagues on bar-chart embellishment and by Bateman and colleagues on visual embellishment show effects that vary by treatment and task, not one law that more ink is worse.

Where it stops holding

Axes, ticks, legends, grouping boundaries, threshold lines, sources, units, method notes, and uncertainty displays are often reading infrastructure. Accessible names, summaries, and data tables also convey necessary information even when absent from the visible plot. Brand or editorial imagery may serve recognition and communication, but must not distort quantities, occlude evidence, or masquerade as data. A null attentional result may reflect an insensitive task or inadequate power.

Applying it

• Assign every element a function: data encoding, reference, grouping, annotation, uncertainty, interaction, accessibility, or communication. Only elements with no task link become removal candidates.
• Prioritise highly salient elements that imply false differences, occlude evidence, or compete with data; do not delete an entire class merely because it is “non-data.”
• Compare critical tasks, delayed recall, and error patterns before and after removal at delivered sizes and devices, reporting uncertainty.
• Preserve provenance, units, state, accessible alternatives, and uncertainty even when they reduce the formal data-ink ratio.

Related

• Same group: U1.14.2 Over-pruning damages readability · U1.14.3 The test is whether removal affects reading
• Adjacent: U1.12.1 A single salient difference can be found in parallel · U1.13.1 Proximity and similarity decide the grouping readers see
• Search terms: data-ink ratio · chartjunk · visual embellishment · attentional competition