Naturalness and aesthetic preference are two important aspects of color rendering that are difficult to capture with rendering indices. Several factors may influence observer's choices in complex ways, e.g. color memory and the composition of the scenes, and the best illumination for specific conditions may be difficult to predict from existing indices.The aim of this work was to estimate psychophysically the spectral composition of the best lighting for commercial food counters. Stimuli were monitor simulations of commercial food counters containing fruits and vegetables derived from hyperspectral data obtained in a local supermarket. Illuminants were synthetized from Judd's daylight spectral basis functions with variable coefficients such that their color defined a chromaticity grid over and around Planckian locus with correlated color temperature (CCT) in the range 2,222—20,000 K. Two conditions were tested: in one, the naturalness condition, observers selected the illuminant producing the most natural colors; in the other, the preference condition, observers selected the illuminant producing the most pleasant appearance.The average CCT in the preference condition was significantly lower than that obtained in the naturalness condition, by about 2,400 K. The average chromaticity of each condition was closer to the Planckian locus than to the daylight locus.