Hierarchy failure shows up as not knowing where to look first

What it is

In a usability session the page appears, people go quiet for two or three seconds, then say “there’s a lot going on,” eyes bouncing among blocks, never pointing at one place and starting work. They can read. There is no unique starting point on the stage. As a phenomenon, hierarchy failure is first seen in those seconds without an origin, not in a later “I didn’t like it” on a questionnaire. A page whose viewers can say “look here first” still has hierarchy. A page they cannot, however many kinds of emphasis it used, has already failed.

Why it happens

A starting point is a peak clearly above the runner-up, so gaze is pulled and speech can name it. Flatten the peak and search becomes unguided patrol: short saccades, many revisits, filler words. Silence is roughly the time until first place is chosen. Treating failure as a taste comment sends the fix the wrong way (new style, new illustration). Treating it as a missing origin, the fix is rebuilding a unique peak. The same symptom can come from one-screen inflation, from flattening without compensation, or from something else — the symptom only reports “no first place.” It does not diagnose which cause.

Studying it

From first frame, time to the first task-relevant utterance or the first stable fixation (dwell in one area past a threshold). Past an agreed line (for example still patrolling at 3 s) counts as no origin. Independent variables can be peak gap and object count. Dependents are latency, filler-word count, and whether a “point first with a finger” lands scatter. Think-aloud lengthens latency; report it apart from silent gaze. Do not let eventual task success cancel a missing origin: many people finish, having groped in the dark for the first seconds.

Where it stops holding

Experts opening a daily workbench carry the origin in motor memory; you will not measure a missing start, which does not mean new users have one. On a page whose job is comparison, a short back-and-forth is not failure; wait until they begin to decide and still cannot say “compare from which side.” Incomplete loads, font flashes and a late ad can fake a missing origin; wait for a stable first screen. High cognitive load (hard copy) also produces silence, but then first fixation has often already landed on a large block — an origin without comprehension, which must be coded apart from “gaze still drifting.”

Applying it

• Put “can they immediately point at one place after entry” into the walkthrough, not only whether it looks good.
• On silence and bouncing gaze, ask first whether a unique peak is missing, then delete blocks or raise one. Do not lead with a coaching animation — it fakes an origin that vanishes when the animation is off.
• A real origin must be both fixated and nameable. A moving pointer hand is not enough.
• Check: record five seconds of the first screen. Two people independently mark “stable landing at which second, on which object.” Long median latency plus scattered landings is no origin. Change one thing: pick the object that should be the origin, increase its difference, decrease the neighbours’, record until median latency drops and landings coincide. Do not add a novice coach layer in the same round; it contaminates observation of the hierarchy itself.

Related

• Same group: F3.08.1 Emphasising everything emphasises nothing · F3.08.2 Flattening weakens hierarchy cues and needs other compensation
• Nearby: F3.05.2 Equal-weight foci fork the flow · F3.07.4 Alignment of visual hierarchy with content priority can be tested with eye tracking or first-look tests
• Search terms: centre of interest · first look · visual hierarchy · scanpath