Color Naming Reflects Optimal Partitions of Color Space – Regier, Kay & Khetarpal (2006)
Use data from pre-industrial societies on color naming systems to argue that naming systems reflect locally optimal divisions of the color space based on the number of color words. Optimality is measured as the degree to witch a category includes similar colors and excludes dissimilar colors based on the CIELAB representation of color space. Use sampling to produce optimal color categories and demonstrate their similarity to attested natural language systems. Show that rotating the color space reduces optimality for most natural language color systems. Does not have an account of why there is so much variability in optimality across color language systems. Proposes that current systems can be viewed as points of change from one optimum to another based on language history and contact.