Spectral reflectance is a key material property and contributor to object appearance. While it has long been known that reflectance in a given wavelength interval correlates strongly with reflectances in neighboring ones, this correlative property has only been exploited implicitly before. The present paper therefore presents a new approach to spectral analysis and synthesis that consists of first deriving a spectral correlation profile and then using it for a direct and full sampling of the spectral and color gamuts corresponding to it. The resulting technique can be used to generate natural-like spectra (or spectra following other, specific correlation properties) and it can also be incorporated into Bayesian models of spectral estimation.