This paper introduces metameric psychometrics, a validation framework for instruments that operate without a latent true score. In place of an oracle it substitutes instrument-computed noise floors – operator variance across cross-family model pairs and source sensitivity under signal-source resampling – so every cohort comparison becomes a falsifiable signal-to-noise test with a computed interval. The framework is realized in the Brand Spectrometer, an open pipeline that reconstructs cohort-resolved, eight-dimensional perception vectors from dated public artifacts by aggregating per-artifact reflections over signal sources; the unit of analysis is the cohort pair. Because cohort metameric variance is the measurement, not error around a true specification, the instrument is validated for measurement properties, never criterion accuracy. A pre-registered battery across two windows shows convergent validity, reproducibility, and four-of-five test-retest reliability. Discriminant resolution depends on the metric: a scale-invariant mean-cosine signal-to-noise abstains, whereas a distribution-level, operator-floored separation between whole cohort clouds resolves the owner cohort from the press in one window (signal-to-noise 3.6, permutation p = .003, triangulated) while abstaining elsewhere. The cohort signal lives in magnitude, not mean shape; the instrument resolves real separations and abstains where none clears. The framework may extend to perceptual constructs measured from public text without ground truth. Includes zharnikov-2026ax-brand-spectrometer.yaml (Paper Spec v0.1.0) – a machine-readable specification of the paper's claims, assumptions, and dependencies. The paper's full machine-first bundle (the SPINE claim/dependency graph and the ONTOLOGY term module) lives in the public repository; see https://github.com/spectralbranding/paper-spec for the standard. This PDF is generated programmatically from that machine-first source under a research-as-repository model.
Dmitry Zharnikov (Mon,) studied this question.