Optimal sets of color receptors and color opponent systems for coding of natural objects in insect vision

Model calculations are used to find an optimal color vision system for the coding of natural objects. The criteria to assess the quality of color vision are (a) discriminability between all colors of a given set; (b) discriminability between nearest neighbors (only the most similar colors of a set); and (c) detectability (color difference between target and background). The colored objects investigated are several sets of flower colors, one set of green foliage colors, and one set of fruit colors. A large variety of hypothetical color vision systems was generated by varying both the wavelength positions of photoreceptors and the weights of color opponent processes used to evaluate the receptor inputs. It is shown that the set of spectral receptor types in flower visiting bees (lambda max = 340, 430 and 540 nm) is close to optimal for the discrimination of several sets of sympatric and simultaneously blooming flower colors, as well as for discrimination of green foliage, but not for fruit coloration. For two sets of objects, the illuminant was varied, which changed the results only marginally. Detectability of flowers against background is likewise optimal. The optimal wavelength positions of photoreceptors are largely independent of the particular color opponent mechanisms used to evaluate the photoreceptor signals. Optimal color opponent systems are all those which comprise two opponent processes with weighting factors differing strongly from one another. The evolutionary implications of these findings are discussed against the background of a recent phylogenetic study which showed that wavelength tuning of insect photoreceptors likely predated the evolution of flower color. (C) 1996 Academic Press Limited