For every pair of colors (j, k), the observers selected a pair of Munsell grays (N-a, N-b) such that the lightness difference matched the color difference in size, and the scaled value of color difference was defined as d(jk) = V-a - V-b. On the basis of these data, where (j, k) are limited in the range that can be matched by d(jk) < 4.0 V, the procedure was presented to define predicted values (d) over tilde(jk) for Munsell colors (J, K) between 4V and 7V directly from Euclidean (d) over cap(jk) between points P-j and P-k in the current Munsell solid. The procedure is more practical than the multidimensional scaling representation. Inter-point (d) over cap(jk) are measured in the units of C in the (H, C) plane and the contributions (d) over cap(jk), of IC and IV differences are assumed to be 1 and 2.3. Precision ofpredictions, RMS = {mean of d(jk) -(d) over tilde(nk))(2)}(0.5), is 0.3 V (0.8 C) for 2-D color differences (V-j = V-k). For the set of data on 3-D color differences used in the present study, RMS = 0.6 V (1.7 C). These were compared with the precision of predictions by Judd, Adams-Nickerson formulae, CIE 1976(L*, u*, v*), and CIE94.