contrast (58, 71, p. 546; 79, 82). Lightness variations are revealed which range in magnitude up to a full value step and more.
The variabilities, or 80 percent ranges, of the data in this table yield some idea of the spread of the individual estimates. The latter suggests, in conjunction with the size of n, that differences of a tenth or two in the means are usually not significant.
The several attributive units are so far from equivalent perceptually, that interattributive comparisons of variability are not valid psychologically; and intercomparisons of variability are best confined to the data of a given attribute. In regard to the influence of background on variability, for the given attribute, there seems to be no indication obvious on inspection. Little more can be said in advance of detailed analysis.
The reader should be cautioned that although the estimates in Table II are average estimates, they are not sufficiently reliable to be taken as standards at their face value. They remain to be smoothed and therefore should not be considered or used as standard data in their present form.
Preliminary Smoothing of Chroma
Preliminary smoothed data were made available by Nickerson in 1938 (60). These results are based on observations with charts on grounds of all three reflectances by about one-quarter of the subjects who eventually contributed to the review. Only the constant-value charts were used and only chroma was smoothed. Nevertheless, these earlier estimates are of considerable interest because they indicate some important adjustments in chroma, and they give the best definition of the ideal Munsell system yet available.
Procedure in smoothing
The samples at each of the seven value levels represented by the 1929 chromatic samples were plotted in the (x,y)-diagram of the I.C.I. coordinate system, utilizing for that purpose the Glenn-Killian data (16). The saturation estimates available at that date were summarized relative to each individual Munsell sample, and were plotted on the same diagram. Thus discrepancies between the 1929 Munsell notation and this 1938 visual consensus were revealed as spatial differences between plotted points at each value level.
Next, representative chroma contours were established by drawing smooth free-hand curves through the scattered Glenn-Killian points.
Wherever such a point was not confirmed by the consensus of visual estimates the contour was correspondingly deflected. Not infrequently the average visual estimate was found to fall in the path of a smooth curve and to justify the drawing of a contour different from that suggested by the unadjusted data alone.
The chroma smoothing was affected by two further considerations both of which involved considerable judgment or freedom of choice on the part of the investigator. In drawing the isochroma contours at the given value level, limited concessions were made to the form of the concentric pattern of curves at that level, taken as a whole, according to the principle that the individual curves must bear a systematic relation to each other. Thus, if the chroma /6 contour has a shape differing from that of chroma /2, the chroma /4 contour must be transitionally related to them.
The other consideration involved the interrelationships between similar chroma contours on adjacent value levels. By drawing the several curves of each value level on a large separate sheet of translucent paper and by properly superposing these sheets, a guide was arranged to this intervalue-level smoothing. Here the changes seemed to be transitional from one level to the next, and especially marked in progressing from the middle levels toward either extreme. The end result of applying these two considerations was the desired tridimensionally smoothed system of iso-chroma contours.
Charted results
Greatly reduced reproductions of the final charts are shown in Figs. 7 to 14. In each figure, the small circles represent the plotted Glenn-Killian points and they bear the 1929 Munsell notations. The heavy elliptoid curves are the smooth iso-chroma contours corresponding to chroma /2, /4, /6, and so on. The origin of these concentric curves is the point representing I.C.I. Illuminant C which approximately represents the achromatic axis of the Munsell solid. Radiating