Alignment of visual hierarchy with content priority can be tested with eye tracking or first-look tests

What it is

Alignment is not “the main title feels big enough” in a review. Take a content ranking written beforehand and match it to what people actually look at first. Eye tracking gives the first-fixation landing; a five-second test gives a spoken first place. Both measure “what was treated as most important,” not whether the page is attractive. On a settings screen or a workbench this is cleaner than voting on “brand feel” for a landing page: a standard first, an observation second, mismatch means not aligned. Alignment without observation is a slogan.

Why it happens

A short presentation cuts off the strategy of reading every word and forces the first choice under the salience heuristic, which is sensitive to visual weight and insensitive to later rationalisation. First fixation is earlier still and less wrapped in language. When the two observations disagree on the same area, spoken report has often mixed in “I know what I should have looked at”; change weight against gaze, not the wording of the questionnaire. The method works only if the content ranking is locked before observation and observers never see that list — otherwise you have measured compliance, not perceptual order.

Studying it

Write reproducible protocol: duration (often 3–5 s), prompt (“tell me what this page is emphasising,” not “please find the setting” — the latter pours a task schema in), how areas of interest are drawn (by object, not by grid), and a pass line (share of people whose first fixation or first spoken name lands on content item one). Eye tracking is for first landing and saccade counts; without a device, masked recall and first click are weaker but usable stand-ins. Materials should include the target and a deliberately misaligned control, so the protocol is known to catch failure rather than always reporting “fairly aligned.”

Where it stops holding

Search tasks and time-limited operations overwrite weight; you have measured hunting, not hierarchy. Experts’ first fixation on a familiar layout lands where memory says; use novices or a shuffled layout. Small-n five-second tests are noisy and cannot be a ship gate on three spoken comments, but they can kill obvious mismatch in draft. Chin-rests, failed calibration, and missing images from an ad blocker bias first landings toward the large colour patches that did load; report whether materials fully rendered.

Applying it

• Ship every critical first screen with a one-page protocol: what content item one is, how many seconds, the exact prompt, the pass line.
• In review, read the protocol result before taste. If “tests are too expensive,” at least do paper five-seconds. Do not fall back to a vote.
• On failure, change weight assignment. Do not change the question so people “get it right more easily.”
• Check: run the same protocol on a control that deliberately makes a secondary item largest. If the protocol passes it, the protocol is invalid — fix the protocol first. On the target, if first fixation or five-second first place on content item one is under the agreed share, it is not aligned. Note what actually came first, change only that object’s channels, run again until the line is met or the content ranking itself is revised.

Related

• Same group: F3.07.1 Visual hierarchy must match content priority · F3.07.7 Hierarchy alignment checks should run on real data, not ideal mockups
• Nearby: F3.04 Scan patterns · A5.07 Attentional capture
• Search terms: first-look test · five-second test · first fixation · eye tracking