Project Atlas Color Evidence Registry v0.1

Purpose

This document begins the empirical foundation of the Project Atlas Perceptual Color Genome.

It does not attempt to teach conventional color theory or recommend palettes. Its purpose is to extract bounded findings from color science, vision science, psychophysics, accessibility standards, and visual-search research, then translate those findings into candidate laws that can eventually support composition analysis and generation.

The governing rule is:

No color principle should be stronger than the evidence supporting it.

Scope of This Pass

This initial pass focuses on six foundational questions:

1. What does a numerical color value actually represent?
2. Why does the same color appear different under different conditions?
3. Which dimensions of color contribute to visual attention?
4. Is luminance categorically more important than chromatic contrast?
5. Can hue alone safely communicate structure or meaning?
6. Are perceptual color spaces sufficiently uniform to support direct design calculations?

The emotional and cultural meanings of color are included only where the evidence can be bounded. Broad claims such as “blue is calming” are not accepted at this stage.

Key Findings

1. Color measurements are observer and condition models, not direct descriptions of subjective appearance.
2. Equal numerical changes do not guarantee equal perceptual changes.
3. Color appearance depends on illumination, adaptation, surroundings, display medium, and spatial context.
4. Luminance contrast is powerful, but chromatic contrast can independently support salience, segmentation, and search.
5. Adding chromatic contrast to weak luminance contrast can accelerate attentional selection under tested visual-search conditions.
6. Color cannot safely carry information alone because observers differ, viewing conditions vary, and hue distinctions can collapse without a lightness or shape difference.
7. Preference and emotion cannot be inferred from hue alone. Lightness, saturation, context, naturalness, experience, and population all alter the response.
8. Perceptually motivated spaces such as CIELAB and Oklab are useful models, not perfect maps of human experience.

Foundational Distinctions

Physical stimulus

A spectral power distribution reaching the eye.

Colorimetric representation

A numerical compression of the stimulus based on a standard observer and specified conditions.

Perceptual appearance

The experienced attributes of the color, including lightness, brightness, colorfulness, chroma, saturation, and hue.

Functional role

What the color does in a composition, such as separating, grouping, warning, attracting attention, encoding state, or shaping mood.

Semantic interpretation

What the observer believes the color means based on context, convention, culture, and prior experience.

These levels are related but not interchangeable.

Evidence

CLR-EVD-001 — Standard color values model an observer

Citation

Commission Internationale de l'Éclairage. Colorimetry, 4th Edition.

Source: https://cie.co.at/publications/colorimetry-4th-edition

Evidence grade

A — International technical standard and synthesis

Bounded finding

CIE colorimetry defines standard observers, illuminants, viewing conditions, tristimulus calculations, chromaticity coordinates, color spaces, and color difference methods.

The standard observer is a population model. It does not assert that all individual observers experience every stimulus identically.

Interpretation

A color token is never simply “the color.” It is a value inside a defined measurement system.

Supports

• CLR-LAW-001: Conditional Color Identity
• CLR-LAW-002: Measurement Is Not Appearance

Generalizability limits

Colorimetric equality is strongest under matched viewing conditions and does not fully model individual observer variation, adaptation, material appearance, or complex spatial context.

CLR-EVD-002 — Tristimulus matches can arise from different spectra

Citation

Commission Internationale de l'Éclairage. Colorimetry — Part 3: CIE Tristimulus Values.

Source: https://cie.co.at/publications/colorimetry-part-3-cie-tristimulus-values-1

Evidence grade

A — International standard

Bounded finding

Different spectral distributions can generate equal tristimulus values and therefore match for a defined observer under defined conditions. This is the basis of metamerism.

Interpretation

Two objects or displays can appear to match in one environment and diverge in another because the physical spectra are not identical.

Composition implication

Cross-device and print-to-screen color consistency cannot be guaranteed by matching a single set of nominal RGB or Lab coordinates without controlling the rendering and viewing environment.

Supports

• CLR-LAW-001: Conditional Color Identity
• CLR-LAW-003: Medium-Dependent Equivalence

CLR-EVD-003 — Appearance requires viewing-condition models

Citation

Commission Internationale de l'Éclairage. The CIE 2016 Colour Appearance Model for Colour Management Systems: CIECAM16.

Source: https://cie.co.at/publications/the-cie-2016-colour-appearance-model-colour-management-systems-ciecam16

Evidence grade

A — CIE technical model

Bounded finding

CIECAM16 transforms tristimulus values into perceptual attribute correlates using viewing-condition-specific parameters.

Interpretation

The need for a color appearance model is evidence against treating XYZ, RGB, or hex values as sufficient descriptions of appearance.

Composition implication

Light mode, dark mode, projected display, printed output, and physical material should not be assumed to preserve the same apparent hierarchy merely because the nominal palette is reused.

Supports

• CLR-LAW-001: Conditional Color Identity
• CLR-LAW-004: Appearance Adaptation

CLR-EVD-004 — CIELAB explicitly separates lightness, chroma, and hue

Citation

Commission Internationale de l'Éclairage. Colorimetry — Part 4: CIE 1976 Lab Colour Space.*

Source: https://cie.co.at/publications/colorimetry-part-4-cie-1976-lab-colour-space-1

Evidence grade

A — International standard

Bounded finding

CIELAB defines coordinates corresponding to lightness and opponent chromatic dimensions and provides methods for calculating color differences.

Interpretation

Hue, chroma, and lightness must be treated as separable variables. “Changing the color” is analytically inadequate because the change may occur primarily along one dimension.

Composition implication

Atlas should record at minimum:

• lightness difference
• chroma difference
• hue difference
• background and adaptation conditions
• spatial extent
• medium and gamut

Supports

• CLR-LAW-005: Multidimensional Color Difference

CLR-EVD-005 — Luminance and chromatic contrast jointly affect salience

Citation

Hardman, A. et al. (2020). Chromaticity- and luminance-driven attentional salience in visual search.

PubMed: https://pubmed.ncbi.nlm.nih.gov/32196068/

Evidence grade

C — Controlled visual-search and electrophysiological study

Bounded finding

When large chromaticity contrast was added to targets with low luminance contrast, the latency of an electrophysiological marker of attentional selection was reduced. The findings indicate that both luminance and chromaticity can contribute to attentional salience.

Interpretation

The common design claim that luminance always dominates hue is too broad. Luminance is highly important, but chromatic contrast can add salience, particularly when luminance contrast is weak.

Composition implication

A more defensible model is:

attentional salience =
  f(luminance contrast,
    chromatic contrast,
    target size,
    background,
    distractor distribution,
    adaptation,
    task)

Supports

• CLR-LAW-006: Combined Contrast Salience
• CLR-LAW-007: Task-Conditional Color Priority

Challenges

• The absolute claim that hierarchy is always carried more strongly by luminance than hue.

CLR-EVD-006 — Salience adapts to the color distribution

Citation

McDermott, K. C., Malkoc, G., Mulligan, J. B., & Webster, M. A. (2010). Adaptation and visual salience.

PubMed: https://pubmed.ncbi.nlm.nih.gov/21106682/

Evidence grade

C — Controlled psychophysical experiments

Bounded finding

Adaptation altered visual-search salience along chromatic and luminance axes. Comparable adaptation effects occurred for multiple color directions and for color distributions resembling natural environments.

Interpretation

Salience is relative to recent visual experience. A vivid color repeatedly used throughout a composition or product may lose some of its exceptional status.

Composition implication

Accent color is a limited resource. Repeated exposure changes the visual baseline against which novelty is detected.

Supports

• CLR-LAW-008: Adaptive Salience
• CLR-LAW-009: Accent Dilution

CLR-EVD-007 — Local context contributes to perceived color

Citation

Hurlbert, A. (2004). Color contrast: a contributory mechanism to color constancy.

PubMed: https://pubmed.ncbi.nlm.nih.gov/14650846/

Evidence grade

B/C — Review and synthesis of psychophysical and physiological evidence

Bounded finding

Local chromatic contrast contributes to color constancy. Texture differences can weaken chromatic contrast induction, while tested relative-motion and relative-depth differences did not produce the same weakening. The reviewed evidence places important contrast mechanisms at early stages of visual processing.

Interpretation

Surrounding colors do not merely decorate a target color. They participate in constructing its appearance.

Composition implication

Color tokens cannot be validated only in a palette sheet. They must be tested inside the intended spatial and textural context.

Supports

• CLR-LAW-010: Contextual Color Construction
• CLR-LAW-011: Palette Non-Independence

CLR-EVD-008 — Web contrast standards measure relative luminance

Citation

World Wide Web Consortium. Web Content Accessibility Guidelines 2.2.

Source: https://www.w3.org/TR/WCAG22/

Evidence grade

A for the normative standard; mixed for universal perceptual prediction

Bounded finding

WCAG defines contrast ratio from the relative luminance of the lighter and darker colors:

(L1 + 0.05) / (L2 + 0.05)

The ratio ranges from 1:1 to 21:1. WCAG also prohibits using color as the only visual means of conveying information in specified contexts.

Interpretation

WCAG contrast is a luminance-based compliance model. It does not quantify every aspect of readability, spatial context, font rendering, adaptation, or chromatic differentiation.

Composition implication

Passing contrast is a necessary floor in many contexts, not proof of complete legibility or hierarchy.

Supports

• CLR-LAW-012: Compliance Is Not Perception
• CLR-LAW-013: Redundant State Encoding

CLR-EVD-009 — Hue preference interacts with lightness and saturation

Citation

Skelton, A. E., Catchpole, G., Abbott, J. T., Bosten, J. M., & Franklin, A. (2017 indexing varies by source). Color preferences in infants and adults are different.

PubMed: https://pubmed.ncbi.nlm.nih.gov/23435629/

Evidence grade

C — Controlled comparative study

Bounded finding

Adults in the tested population commonly preferred blues and least preferred greenish yellows, but hue preference interacted with lightness and saturation. Infant preferences differed from adult preferences.

Interpretation

Preference is not a stable lookup table from hue to emotion. Development, experience, saturation, and lightness alter the result.

Composition implication

Brand and emotional color claims should be represented as population- and context-dependent probabilities rather than universal mappings.

Supports

• CLR-LAW-014: Conditional Color Preference

Challenges

• Universal “color psychology” charts.
• Hue-only emotional classifications.

CLR-EVD-010 — Naturalness affects preference for color compositions

Citation

Nascimento, S. M. C. et al. (2021). Preference for color compositions perceived as natural.

PubMed: https://pubmed.ncbi.nlm.nih.gov/33965779/

Evidence grade

C — Controlled image experiments

Bounded finding

Manipulating image color gamuts changed perceived naturalness and preference. The study supports a relationship between ecological plausibility and the evaluation of color compositions.

Interpretation

People may prefer some palettes not because of geometric relationships on a color wheel, but because the palette resembles learned regularities in natural scenes.

Composition implication

“Color harmony” may partly emerge from statistical familiarity, semantic coherence, and ecological expectation rather than hue-angle geometry alone.

Supports

• CLR-LAW-015: Ecological Color Coherence

Challenges

• The assumption that fixed complementary, analogous, or triadic geometry is a complete explanation for harmony.

CLR-EVD-011 — Oklab is a practical perceptual model, not settled ground truth

Citation

Ottosson, B. (2020). A perceptual color space for image processing.

Source: https://bottosson.github.io/posts/oklab/

Evidence grade

E/C — Engineering model supported by comparative datasets and tests, but not a formal international standard

Bounded finding

Oklab was designed to improve practical prediction of perceived lightness, chroma, and hue while remaining computationally simple for image processing and interpolation.

Interpretation

Oklab and OKLCH are useful working spaces for design systems because their dimensions are easier to manipulate than raw RGB. They should not be treated as perfectly perceptually uniform.

Composition implication

Atlas may use OKLCH for implementation and token generation while retaining CIE color-difference methods and empirical validation for research claims.

Supports

• CLR-LAW-005: Multidimensional Color Difference
• CLR-LAW-016: Model-Bounded Uniformity

Observations

CLR-OBS-001 — A palette is a system of relationships

Observation

The evidence does not support evaluating a color independently of its background, adaptation state, medium, or neighboring colors.

Interpretation

A palette entry should be stored as a node with conditional relationships, not as an isolated hex value.

Confidence

High

CLR-OBS-002 — Luminance is foundational but not sufficient

Observation

Luminance supports boundary detection, contrast measurement, and readability, but chromatic contrast can independently alter attention and segmentation.

Interpretation

Atlas should reject both extremes:

• hue is enough
• only luminance matters

Confidence

Moderate to high

CLR-OBS-003 — Accent power depends on rarity

Observation

Adaptation and visual-search research indicate that salience depends on the distribution of features in the surrounding field and recent experience.

Interpretation

An accent color used everywhere stops functioning as an accent.

Confidence

Moderate

CLR-OBS-004 — Accessibility and hierarchy are different questions

Observation

A color pair can satisfy a luminance contrast rule while still failing as a complete state, grouping, or attentional system.

Interpretation

Atlas must distinguish:

1. detectability
2. readability
3. identifiability
4. grouping
5. state interpretation
6. attentional priority

Confidence

High

CLR-OBS-005 — Harmony may be ecological as well as geometric

Observation

Preference can depend on naturalness and familiar environmental color statistics.

Interpretation

Color-wheel geometry is at best one contributor to perceived harmony.

Confidence

Moderate

Candidate Laws

CLR-LAW-001 — Law of Conditional Color Identity

Hypothesis

The perceived identity of a color depends on the stimulus, observer, illumination, adaptation, surrounding field, spatial scale, and medium.

Prediction

A fixed colorimetric or RGB value will receive different appearance matches or attribute ratings when one or more contextual variables change.

Supporting evidence

• CLR-EVD-001
• CLR-EVD-002
• CLR-EVD-003
• CLR-EVD-007

Confidence

High

CLR-LAW-005 — Law of Multidimensional Color Difference

Hypothesis

Color difference cannot be adequately represented as unweighted distance in device RGB coordinates.

Prediction

Pairs with equal RGB distance will not produce equal perceived difference, while perceptually motivated color-difference models will improve prediction under their validated conditions.

Supporting evidence

• CLR-EVD-004
• CLR-EVD-011

Confidence

High for the inadequacy of raw RGB distance; moderate for any one replacement model

CLR-LAW-006 — Law of Combined Contrast Salience

Hypothesis

Attentional salience can be produced by luminance contrast, chromatic contrast, or their interaction.

Prediction

Adding task-relevant chromatic contrast to a low-luminance-contrast target will improve selection speed or accuracy under at least some visual-search conditions.

Supporting evidence

• CLR-EVD-005
• CLR-EVD-006

Counter evidence and boundary

The relative contribution will vary with target size, spatial frequency, eccentricity, background, color direction, adaptation, and task. This law does not imply that hue can replace adequate text luminance contrast.

Confidence

Moderate to high

CLR-LAW-008 — Law of Adaptive Salience

Hypothesis

The salience of a color feature decreases as the visual system adapts to that feature distribution.

Prediction

A color initially detected rapidly as an outlier will lose search advantage as the same or similar color becomes frequent or repeatedly viewed.

Supporting evidence

• CLR-EVD-006

Confidence

Moderate

CLR-LAW-010 — Law of Contextual Color Construction

Hypothesis

The apparent color of an element is partly constructed from local and global context rather than determined solely by the element's physical or numerical value.

Prediction

Holding the target value constant while changing surrounding color, texture, or adaptation will alter appearance judgments.

Supporting evidence

• CLR-EVD-003
• CLR-EVD-007

Confidence

High

CLR-LAW-012 — Law of Compliance–Perception Separation

Hypothesis

Passing a formal contrast threshold does not guarantee successful recognition, grouping, state interpretation, or attentional priority.

Prediction

Some displays that pass a contrast requirement will still show performance differences when typography, crowding, adaptation, glare, chromatic confusion, or competing hierarchy is manipulated.

Supporting evidence

• CLR-EVD-008
• Existing Project Atlas crowding and recognition evidence

Confidence

High as a conceptual boundary; quantitative UI thresholds remain to be derived

CLR-LAW-013 — Law of Redundant State Encoding

Hypothesis

Critical information encoded by color is more robust when reinforced by an independent cue such as shape, text, position, pattern, or iconography.

Prediction

State-identification accuracy across varied observers and viewing conditions will be higher for redundant encodings than for hue-only encoding.

Supporting evidence

• CLR-EVD-008
• Existing Project Atlas Reinforced Structure and Cue Competition laws

Confidence

High as an accessibility principle; implementation effectiveness depends on cue compatibility

CLR-LAW-014 — Law of Conditional Color Preference

Hypothesis

Color preference is a function of hue, lightness, saturation, context, development, experience, and population rather than hue alone.

Prediction

Preference rankings will change when lightness, saturation, surrounding colors, image semantics, population, or task changes while hue is held constant.

Supporting evidence

• CLR-EVD-009
• CLR-EVD-010

Confidence

Moderate to high

CLR-LAW-015 — Law of Ecological Color Coherence

Hypothesis

Color combinations that preserve familiar environmental or semantic relationships will often be judged more natural and may be preferred over equally structured but ecologically implausible combinations.

Prediction

Rotating or remapping an image gamut while preserving spatial structure will reduce naturalness and preference when the transformation violates learned color regularities.

Supporting evidence

• CLR-EVD-010

Confidence

Moderate

CLR-LAW-016 — Law of Model-Bounded Uniformity

Hypothesis

Every perceptual color space is an approximation whose uniformity depends on the dataset, task, gamut, adaptation, and color-difference scale used for its construction or evaluation.

Prediction

A color space that performs well for interpolation or moderate sRGB differences will show systematic errors in at least some other hue, chroma, adaptation, or suprathreshold conditions.

Supporting evidence

• CLR-EVD-004
• CLR-EVD-011
• The continuing development of revised CIE color-difference and appearance models

Confidence

High

Provisional Color Relationship Model

Observed Color Outcome =
    Physical Stimulus
  × Observer Sensitivity
  × Adaptation State
  × Viewing Conditions
  × Spatial Context
  × Temporal Context
  × Task Relevance
  × Learned Meaning

A more explicit placeholder:

Cₒ = f(S, O, I, A, B, X, M, T, R, K)

Where:

• S = spectral or encoded stimulus
• O = observer characteristics
• I = illumination
• A = adaptation state
• B = background and neighboring colors
• X = spatial extent and position
• M = medium, material, and display
• T = temporal exposure and change
• R = task relevance
• K = learned semantic and cultural knowledge

This is a causal inventory, not a fitted equation.

Data Schema for Future Evidence Extraction

evidence_id:
citation:
year:
discipline:
evidence_grade:
genome_nodes:
population:
observer_characteristics:
sample_size:
stimulus_medium:
display_or_material:
illuminant:
adaptation:
viewing_distance:
visual_angle:
background:
spatial_context:
task:
independent_variables:
dependent_variables:
color_space:
color_difference_formula:
quantitative_results:
effect_size:
statistical_significance:
authors_conclusion:
bounded_finding:
atlas_interpretation:
candidate_laws_supported:
candidate_laws_challenged:
generalizability_limits:
replication_status:
source:

Open Questions

1. Under which spatial-frequency and target-size conditions does luminance dominate chromatic contrast?
2. How does chromatic contrast interact with peripheral crowding?
3. What is the best practical perceptual space for UI token generation versus scientific color-difference prediction?
4. How quickly does repeated accent use reduce salience?
5. How should color hierarchy be modeled in mixed light and dark surfaces?
6. How do age, lens yellowing, color-vision deficiency, glare, and low vision alter usable chromatic separations?
7. Can palette harmony be predicted from natural-image statistics, semantic relationships, or processing fluency?
8. When do multiple colors cease to form hierarchy and begin to create competition?
9. How does colored area affect apparent lightness, chroma, and dominance?
10. Which findings transfer from isolated laboratory patches to complex interface, editorial, architectural, and artistic compositions?

Next Actions

1. Build a dedicated evidence track for luminance versus chromatic contrast.
2. Extract quantitative visual-search results by target size, eccentricity, and distractor set size.
3. Build an accessibility track covering normal vision, color-vision deficiency, aging, glare, and low vision.
4. Compare CIELAB, CIEDE2000, CAM16-UCS, Oklab, and OKLCH by intended use.
5. Build a contextual appearance track covering simultaneous contrast, adaptation, induction, and color constancy.
6. Build a preference and emotion track that explicitly records population, object context, saturation, lightness, and semantic meaning.
7. Link these laws to existing Atlas laws concerning similarity, cue competition, search competition, reinforced structure, and recognition beyond visibility.

Revision History

Agent Instructions

When creating or modifying this document:

1. Separate observation from interpretation.
2. Never strengthen a conclusion beyond the available evidence.
3. Preserve contradictory findings.
4. Prefer measurable variables over subjective descriptions.
5. Reference candidate laws and genome nodes whenever possible.
6. Use stable IDs for observations, evidence, laws, experiments, metrics, and case studies.
7. Record assumptions explicitly.
8. Record confidence explicitly.
9. Keep the YAML header valid.
10. Do not delete revision history; append to it.