Page:Journal of the Optical Society of America, volume 33, number 7.pdf/64

 5PB; between 5P and 7.5P; and near 5G, probably toward 10GY. The averages of these hue lines cut the spectrum locus at 572.2, 505, 473.5 mμ, and the purples at 559c. Purdy reported invariables at 571, 506, and 474 mμ and Exner at 577, 508, and 475 mμ. Evidence of this changing relation between hue and dominant wave-length is found in the earliest Munsell papers [(8), Fig. 5], and the observations on which this report is based (28) emphasize it.

As a precautionary measure, it seems well to emphasize the fact that the establishment of standard hue loci was a difficult problem. Considerable judgment and inference were necessarily relied upon, and the complete result is not to be regarded as more than a first approximation to uniform hue spacing. The problem was especially difficult at the 1/ and the 9/ value levels to which it seemed very desirable to try to extrapolate but at which there were no chromatic Munsell samples. Such extrapolation is hazardous and tentative as sudden changes may readily occur in these extreme regions of the solid; however, it seemed best to make a beginning and thus provide some basis for future refinement.

After the 20 principal hue lines had been established at all value levels, the final step of drawing in 20 more intermediate lines to correspond approximately to the recently released third intermediate hues was a simple one. The latter 20 were drawn in simply to fit into the sequences of the original 20 which are based upon the visual estimates of the 1929 samples. The interpolated 20 fit the corresponding data fairly well. This latter group of hue lines was plotted in a figure—not shown—similar to Fig. 13. The entries in Table I were read from large master charts on which the extrapolated hue and chroma lines had been carefully traced. Then check readings were made from the originals of the charts shown in Figs. 1-9. The entries were read from the original penciled charts before inking.

The tables and charts of the chroma and hue loci were worked out so that they would be correct for the new value-reflectance relations given in Table II. The procedure for determining the latter is described in the next section.

The hue and chroma adjustments were made on the basis of visual estimates which had been averaged for all three observing grounds. This average was taken because under our conditions of observation the influence of background reflectance on hue and saturation comparisons was neither systematic nor dependable, and is therefore to be regarded as insignificant. The influence of background reflectance on value estimates, however, was found to be significant. The general nature of this effect is evident from intercomparison of the three curves in Fig. 14 which correspond to the observations on the white, gray, and black grounds. In some of the more extreme cases encountered in practice it will be worth while to make an adjustment for this effect of simultaneous contrast, and this can be done by reference to the corresponding graph. In general and for intermediate reflectances, however, it seemed desirable to standardize on a single value function suitable for use with a light ground, otherwise complications would arise in the hue and chroma dimensions. This decision was supported further by the discovery that only the function for the black background was greatly out of line.