Perceptual and neural constraints on photometric measures of heterochromatic brightness

Luminance and the candela provide the foundation of photometry and define standard measures of the visual impact of light. They are defined by classical photometric methods that fail to capture key aspects of brightness judgments across different colors in spatially extended, real-world scenes: Highly saturated colors appear brighter, mixtures are subadditive relative to luminance, and short wavelengths contribute more strongly than expected. Despite these long-standing discrepancies, progress has been limited by the lack of scalable methods to measure heterochromatic brightness under steady viewing. Here, we introduce a ranking-based paradigm that enables reliable measurement of brightness across large stimulus sets. Observers ranked 144 colored stimuli spanning hue and intensity, yielding highly consistent results across sessions, observers, and display environments, including large-scale online studies. This approach reveals a stable global structure of heterochromatic brightness preserved across devices and viewing contexts. Using these data, we evaluate a broad range of candidate models. Luminance, radiance, and established color appearance models leave key systematic patterns unexplained. In contrast, simple nonlinear pooling rules, specifically taking the weighted maximum of the red, green, and blue channel values, provide a near-ceiling account of observed rankings. Neural recordings reveal a corresponding temporal dissociation: High-frequency responses track luminance, whereas low-frequency responses align with nonlinear predictions. Experiments with spectrally tunable illumination confirm the same pattern in spatially extended scenes. Together, these results establish a scalable framework for measuring heterochromatic brightness and identify a simple computational rule governing brightness under steady viewing.