Metamerism Explained

A metameric match is a conditional agreement: spectrum A and spectrum B collapse to nearly the same color under one illuminant, observer, and geometry, then separate when those conditions move. This lab exposes the moving parts at once: reflectance shape, illuminant SPD, observer shift, surface flare, and pass/fail tolerance.

Two stimuli are metamers when they have different spectral power distributions but produce identical tristimulus values for a given observer under a given illuminant. They look like the same color in those conditions. Change conditions - swap the illuminant, swap the observer - and the stimuli may produce different tristimulus values, and look like different colors.

The phenomenon is unavoidable, not pathological. It is the natural consequence of how human vision works: three cone classes integrate every incoming spectrum down to three numbers. Many spectra map to the same triple. When those spectra are viewed under conditions that change the integration, the agreement breaks down.

Most painted matches in the real world are metameric matches, not spectral matches. The dye chemist did not reproduce the original spectrum - they just produced a spectrum that integrates to the same XYZ under the agreed reference illuminant. That's why a "color match" on a spec sheet always specifies the illuminant and observer it was measured under.

Take a spectral power distribution sampled every 5 nm from 380 to 780. That's 81 numbers - an 81-dimensional point. Integration against the three color matching functions x̄, ȳ, z̄ projects that 81-dim point down to three numbers. By a basic linear-algebra argument, the projection has a kernel: spectral changes that lie in the kernel produce no change in XYZ.

The kernel of M has dimension 78. That means for every visible color, there is a 78-dimensional family of spectra that produce identical XYZ. Add any vector from this null space to a known SPD, and the XYZ is unchanged. The two spectra are metamers by construction.

In practice not every spectrum in the null space is a physically realizable reflectance - reflectance has to stay in 0..1, dyes have characteristic absorption shapes, and so on. The 78-dim space is enormous but constrained. Even so, multiple physically achievable spectra typically exist for any given target XYZ, which is what makes metameric matches commercially possible.

Metamerism is not one phenomenon. Changing different aspects of the viewing situation produces different kinds of mismatch. Four are commonly distinguished.

The first two are the workhorses. The remaining sections of this article look at each kind in detail, with demos and the standards used to quantify them.

Two surfaces with different reflectance spectra R₁(λ), R₂(λ) can be made to integrate to the same XYZ under illuminant S₁(λ). The product R₁(λ) · S₁(λ) and R₂(λ) · S₁(λ) integrate to identical triples. Now replace the illuminant with S₂(λ) - the products become R₁(λ) · S₂(λ) and R₂(λ) · S₂(λ). These are different products. Their tristimulus values are no longer guaranteed to match. Usually they don't.

The CIE 1931 standard observer is an averaged set of color matching functions. Real cone sensitivities vary from person to person. Two people whose L-cone peaks are 5 nm apart will integrate the same spectrum to subtly different tristimulus values. A match for one might fail for the other - particularly when the two metameric spectra differ around wavelengths where their cones disagree.

Observer metamerism used to be a niche concern. It has become urgent with narrow-band displays. OLED phones, laser projectors, and HDR cinema use primaries with very narrow spectral peaks - exactly the conditions that amplify observer disagreement. Two observers can look at the same display and disagree about whether the screen and a reference sample match, where 20 years ago on a broad-band CRT they would have agreed.

The CIE has two standard observers: 1931 2° and 1964 10°. The 2° observer is built for small visual fields - the kind that fit entirely within the cone-dense fovea. The 10° observer handles larger fields that extend into the cone periphery, where rod and cone densities and spectral sensitivities differ.

A metameric pair calibrated for the 2° observer may not match for a 10° observer. The change is most pronounced for samples with sharply peaked spectra - the kind produced by tri-band fluorescents and some LEDs. Paint chips viewed across a tablet, fabric samples held next to a finished garment, food on a display vs in a bowl - all involve different angular sizes and therefore different observer choices.

Spectrophotometers measure samples at fixed geometry - typically 45°/0° (illumination at 45°, viewing at 0°) or d/8 (diffuse illumination, viewing at 8°). The reading tells you about the spectrum at that geometry. In real life, surfaces are viewed from many angles, and surfaces with appreciable gloss, sparkle, or angular dependence produce different spectra at different geometries.

Two samples can match perfectly under instrument geometry and then look different in a casual hand-held comparison. Metallic paints are notorious: the same target metallic and a matched non-metallic look identical face-on and disagree dramatically when tilted. Pearlescent finishes, fluorescent dyes, and brushed metals all carry the same risk.

When a metameric pair matches under reference illuminant A and a measurement is made under test illuminant B, the resulting color difference is the special index of metamerism M_I,t. It is simply the ΔE between the two samples computed under the test illuminant.

Lower is better. A perfect spectral match has M_I = 0 under every test illuminant. A poor metameric match can have M_I = 5 or more, meaning a clearly visible shift between the samples under the test condition. Painted and dyed products typically specify maximum allowed M_I across a list of test illuminants (A, D65, F2, F11) as part of the quality contract.

Two ways to make two samples the same color. A spectral match uses exactly the same reflectance curve, point for point. Light hits them the same way under every illuminant, scatters into every observer's eye the same way, and is judged identical regardless of conditions. A metameric match produces the same tristimulus under one illuminant and observer, with no guarantee about any other condition.

Metamerism cannot be eliminated. It can be detected, measured, and mitigated. Most of the production-side discipline of color science is built around this.

Six questions on metamerism, observers, and illuminant dependence. Wrong answers come with brief explanations.