First impressions: novel yet recognizable

Four homepages with identical content are rated after 50 ms and 500 ms exposures.

005 · Project presentationInsufficient evidenceInsufficient evidenceStarted 2026-10-03 · updated 2026-10-03

Demo width

Full768375

Loading demo…

Principles

A work index is judged almost instantly; its initial view should convey what the site contains and its quality without requiring interaction.

EmpiricalLindgaard et al. (2006)Appeal ratings after 50 ms exposure strongly correlated with ratings after 500 ms.
EmpiricalReinecke et al. (2013)First impressions can influence subsequent usability and trust judgments.

Novel, yet recognizable

Attractive websites balance novelty with familiarity for their type; novelty may matter more for service sites and typicality for commercial sites.

EmpiricalSilvennoinen et al. (2026)A survey of 108 participants across six commercial and six service websites found that novelty and typicality both predicted appeal (the MAYA principle), with novelty the stronger predictor.

Hypothesis

If four content-identical homepages are assessed after 50 ms and 500 ms, (a) appeal rankings should be similar across exposure durations, and (b) a more novel stimulus (S4) should be rated more appealing than the conventional stimulus (S1) without reducing the ability to identify the site's content.

Measured by: Two-block protocol B with randomized stimulus order within each block. Block 1: 50 ms exposures followed by 1–9 appeal scores. Block 2: 500 ms exposures followed by 1–9 appeal, 1–7 novelty, 1–7 typicality, and multiple-choice content identification. Each exposure is measured frame by frame and marked if its duration deviates by more than 10%.
Falsified if: Falsified if (a) 50 ms and 500 ms appeal rankings are not positively correlated, or (b) S4 is not more appealing than S1 at 500 ms, or (c) S4 has lower content-identification accuracy than S1.

Implementation

S1 · S1 · Conventional: Typical homepage: top navigation, left hero, right-side numbers, cards below.
S2 · S2 · Polished conventional: A familiar structure using the lab's styling and area tints.
S3 · S3 · Typographic index: A large-typography experiment list as the dominant element.
S4 · S4 · Research chain diagram: The lab's research method represented as a node-and-line diagram.

Four 1280×800 JSX stimuli use identical content (brand, heading, subtitle, navigation, counts, four experiments) with no images. Block 2 is reshuffled, so participants see each stimulus twice; that is a documented design limitation, not a controlled variable.

Appeal, novelty, typicality, and content questions are draft items: the full Silvennoinen et al. wording is paywalled, and VisAWI-pos had only been validated in English, so it was not used in the original Indonesian protocol B.

Calibration measures the median of 40 frames when the demo opens and warns when 50 ms cannot be achieved. Every exposure is timed frame by frame; deviations beyond 10% are flagged in the table, counted in the summary, and invalidate the corresponding measurements.

The first instrument unmounted the stage during questions, so the second exposure encountered a null ref and aborted. An automated probe found the issue; it was fixed by keeping the stage mounted but hidden while answering, which also ensures the next stimulus is ready.

TSV export uses navigator.clipboard with a textarea fallback. The fallback was verified when clipboard access was denied in headless tests; the primary clipboard path was not verified in a real browser.

Observations

Timing instrument verification

Run the complete sequence of eight scripted exposures, answering appeal 7, novelty/typicality 4, and the first content choice. Three runs at 1440 px and one at 375 px; ranges cover all runs and 16 exposures per target.

Chrome 154 headless, 30 Hz (batas headless di mesin ini), 2026-10-03, 1440×900 and 375×812

Target	Measured	Frames	Flagged ⚠
50 ms	33.4 or 66.6–66.7 ms	1–2 frame	16 of 16
500 ms	499.9–500.1 ms	15 frame	0 of 16

At 30 Hz, a 50 ms exposure cannot be timed exactly: the result is approximately 33.4 ms (one frame) or 66.7 ms (two frames), depending on frame phase. All 16 instances were correctly flagged. All sixteen 500 ms exposures took fifteen frames and met their timing target. Stimulus order differed between blocks and runs; all answers were captured.

Unstable frame rate, reduced motion, and narrow width

Run once without a frame-rate cap (around 56 Hz, unstable); once at 1440 px with reduced motion emulated; and inspect the narrow-width warning at 375 px.

Chrome 154 headless, 2026-10-03

Condition	50 ms exposure	500 ms exposure
Unstable rate ±56 Hz	41.6–57.3 ms, out-of-tolerance marked	492.7–508.1 ms, 0 flagged
Reduced motion	Same as normal run; reduced-motion label shown	499.9–500.0 ms, 0 flagged
375 px width	Desktop-required warning shown; no overflow	499.9–500.1 ms, 0 flagged

With unstable frame timing, 50 ms exposure duration also fluctuates and only out-of-tolerance values are flagged; median calibration does not guarantee every exposure. Reduced motion does not change stimulus exposure, and the narrow-width warning appeared correctly.

Findings

Insufficient evidence

The pilot confirmed its procedure: 500 ms exposures were timed accurately, unachievable 50 ms exposures were flagged, orders were randomized, and answers were recorded. But no human participants have taken part, so appeal rankings at 50 ms versus 500 ms cannot be compared and the hypothesis remains undecided.

Limitations

Zero participants; every answer came from a scripted probe.
Appeal items remain drafts; their wording was not taken from the original publication.
50 ms exposures were never timed accurately: headless Chrome ran at 30 Hz on this machine and uncapped timing fluctuated.
Headless Chrome only; Safari, Firefox, and 120 Hz displays were not tested.

Insufficient evidence

As with the paired principle, the pilot shows that the demo can compare stimuli S1 and S4, but without human participants there are no comparable judgments of novelty, typicality, or appeal.

Limitations

Zero participants; every answer came from a scripted probe.
Novelty, typicality, and content items remain drafts.
S4 is almost certainly more visually complex than S1–S3; equal complexity has not been achieved or measured from screenshots.
Block 2 repeats the same stimuli, so judgments may be influenced by Block 1.

Evaluation

Usability: One Start button; eight exposures followed by twenty scale/multiple-choice responses; Escape cancels; results plus a Copy TSV button. The sequence can be completed without explanation.
Accessibility: Use number keys 1–9 for scale responses; questions use an aria-live region; visible focus uses the global outline; fixation, stimuli, and mask are aria-hidden. Reduced motion does not alter exposures. DOM checks found named buttons; screen language was previously Indonesian. axe-core was not run.
Responsive: 1280×800 stimuli scale to screen height; below 1024 px a warning says the pilot requires desktop. A full 375 px run finished without page overflow.
Performance: The demo transferred about 159 KB of JavaScript. Only layer opacity changes during exposures; stage height is scaled to avoid layout shift.
Craft: Layers swap on the same frame; each stimulus is already mounted at opacity 0. Calibration, warning flags, and TSV export make outputs inspectable; decimal formatting is being localized to English.

Area: Project presentation