Shadows, 3-D bevels, and decorative animation make the eye work harder for no judgment gained
Aliases: chartjunk in operational displays · control-room interface
What it is
Chartjunk covers shadows, textures, 3-D perspective, background illustration, and semantically empty animation — visual features that support no state judgment at all. The harm goes beyond making a display look cluttered: the more specific problem is that operators can misread these features as a form of data encoding, taking a purely stylistic shadow as a hierarchy or anomaly cue, or a decorative gradient as an implied severity level.
Why it happens
Visual search is fundamentally about picking a real target out of a field of features, and search speed depends on whether the target can pop out against the distractors. Decoration is worse than simply adding one more irrelevant element, because many decorative techniques generate strong edges, contrast, or motion transients — a shadow adds an edge, a gradient adds contrast, an animation adds a motion signal — and these are exactly the feature types that a viewer's bottom-up salience mechanism is drawn to before any conscious judgment occurs. That means the cost of decoration is not simply "one more glance" added on top; it can actively redirect the very first fixation toward the decoration instead of the actual anomaly, a systematic, directional cost rather than a random increase in search time. Three-dimensional perspective carries a more specific problem: perspective rendering systematically distorts perceived area and position, so two values that are actually close can look very different, or vice versa, once rendered in 3-D. Shadows can likewise imply a hierarchy or device state that does not exist, making an element look "raised" and therefore actionable, or making a purely stylistic highlight look like an anomaly indicator.
Where it stops holding
Not every nondata visual mark is decoration. Lines and whitespace that provide grouping, indicate connection, show orientation, or signal interactivity are functional structure, and removing them genuinely harms task performance — they should not be lumped in with pure shadows, textures, or 3-D effects. The distinguishing test is concrete: does removing this visual element make the information operators need to complete the task worse? If it does, it was not decoration; if nothing changes, it was. The goal is not an interface with no style at all, but one where every visual feature maps to a function that can be stated explicitly.
Applying it
Before removing any visual element, keep a comparable baseline prototype so a subtle grouping cue that experienced operators actually rely on is not mistaken for decoration and deleted along with it — much of what seasoned operators depend on is exactly this kind of unobtrusive cue. Annotate, layer by layer, the specific data meaning, interaction function, or grouping purpose of every graphical element on the display, and remove anything for which no purpose can be stated. How to check: on real panels under real lighting, test the before-and-after versions for time to locate the first anomaly and misreading rate, and record or observe the change in scan path, rather than relying on a subjective sense that the interface now "looks cleaner."