EX-ITTEN-001 — Extension ratios, area, and salience

Question and falsifier

Do fixed color-area ratios predict perceived balance after luminance, chroma, position, and semantic priority are controlled? The fixed-ratio account fails if those variables explain balance and attention better than complementary area ratio, or if the preferred ratio changes materially by task or observer.

Work completed

A deterministic full-factorial pipeline test crossed four area fractions, four luminance differences, three chroma differences, three positions, and two semantic priority levels: 288 modeled conditions. The proxy was intentionally explicit:

area^0.62 × hypot(3.2ΔL, 1.7ΔC) × center-bias × semantic-priority.

This tests design, data shape, reproducibility, and analysis sensitivity—not people.

Pilot results

Proxy-score correlations were area r=.5161, luminance difference r=.7096, chroma difference r=.0704, and semantic priority r=.2286; score range was 0.0152–1.2586. Because the generating model assigned those weights, these are successful pipeline-recovery results, not support for the coefficients. The run demonstrates why area alone cannot identify “visual weight.”

Human experiment ready to run

• Design: within-participant 4 × 4 × 3 × 3 factorial subset selected with D-optimal balancing; include Itten ratios plus logarithmically spaced alternatives.
• Stimuli: calibrated, texture-free pairs and complete interface compositions; match or independently manipulate CAM16 lightness and chroma.
• Outcomes kept separate: two-alternative balance judgment, adjustment-to-balance, first fixation, dwell distribution, target-search accuracy, and preference.
• Participants: initial 60 normal trichromats plus 30 participants with diagnosed red–green CVD; expand after pilot variance and reliability estimates.
• Analysis: mixed-effects psychometric model with participant and stimulus random effects; compare area-only, feature, and feature-plus-semantic models by held-out likelihood. Do not treat gaze as balance.
• Controls: randomized side, gaze-calibrated center fixation, equal exposure, reading direction, visual acuity, age, color-vision assessment, and device calibration.

Current position / continuation

Stage: computational pipeline complete; human evidence absent. Next action: create calibrated stimulus set, preregister model comparison, obtain ethics/participant consent as locally required, and collect the first 15-participant reliability pilot. Data: ../experiments/data/ex-itten-001.json. Code: scripts/itten-color/run-computational-pilots.mjs.