Color differences are almost always described by ΔE in Lab color space. This space, defined by the 1976 CIE report, is calculated using the Tristimulus values X, Y, Z, defined in the CIE 1931 report. Further, a complex image is often evaluated by averaging the individual ΔEs to calculate a Color Metric for the color difference between two images.The experiments in this paper generate triplets of images: one is defined as “Original” the other two as “Reproductions.” Each area in the “Reproduction” differs from the “Original” by a constant ΔE. The goal of the experiments is to see if some choices of colors make better reproductions than others.The results show that color metrics comparing color differences across edges within the same image predict better reproductions than color metrics comparing absolute Lab values of corresponding areas in different images.