a close approximation to the psychological solid is that the resulting scales facilitate adequate interpolation among the adjacent members of a finite system of samples. For if the chromatic continua are equi-stepped variations in but one attribute at a time, an unknown color can be evaluated quite definitely and quickly by assigning to it the notational equivalent of its appropriate position in the color system (39, p. 362). In the absence of such psychological scales, on the contrary, a relatively haphazard and laborious succession of comparisons must be made. That is because close interpolation is then scarcely possible, and the system must include a far larger assortment of samples to permit the employment of the alternative method of evaluating the unknown by matching or approximately matching it.
A further practical advantage of uniform colorscales lies in the convenient assignment of color tolerances. Of course, “noticeability of color variations” is not the only important determinant of color tolerances (48, p. 416). On the other hand, the general population considers that visual appearance or “how it looks” is very often crucial, and so perceptual tolerances are correspondingly important for many scientific and industrial colorists (5, 9, 27, 35, 36, 39, 57, 59).
Spacing Problem and Procedure
Historical note
Long recognized as the outstanding practical device for color specification by pigmented-surface standards (38), the Munsell system of color (49, 52-54) has become the subject of several other recent studies (15, 16, 23, 61). As early as 1919, however, I. G. Priest, Gibson, and others were already making constructive proposals for its improvement with special emphasis on scientific specification and the value scale (68) of the Atlas colors (54). Subsequently, A. E. O. Munsell, Godlove, and Sloan markedly improved the neutral-scale value spacings (18, 50). When the original report of the Glenn-Killian data (16) was circulated in 1935, numerous irregularities became apparent. There were general irregularities which could be referred to the I.C.I. coordinate system in which the data were expressed, and there were local irregularities due to clerical or to measurement errors, as one might expect. A third class of irregularities could be ascribed to various field factors including spatial arrangement and background reflectance, but a fourth class seemed to be due to real errors in the Munsell samples themselves (49). Some of these irregularities are shown in Figs. 7 to 14, to be discussed later.
The idea of improving the Munsell system by visually smoothing it appears to have occurred independently to three people in 1935, viz., H. P. Gage, Dorothy Nickerson, and W. B. VanArsdel. In 1936 there appeared a mimeographed statement by D. B. Judd and Dorothy Nickerson entitled the “Review of the spacing of the Munsell colors” (40), which not only pointed out the desirability of smoothing irregularities but also outlined procedure for reviewing the constant-value charts and recording the observers’ estimates of the true notation for each sample. The procedural details employed in securing most of the present data will be described later. Suffice it to say here that the principal features were included in that original statement of 1936 and that the essential ratio method involved had been known as early as 1929 (73).
The smoothing process is designed, of course, to eliminate only those irregularities which represent real departures from psychological regularity in the samples themselves. Thus sweeping irregularities in the coordinate system of reference can be allowed to remain as a part of the normal baseline. Most of the clerical errors can be corrected by checking. Irregularities due to field factors can be controlled, in part at least, by the systematic employment of masks and backgrounds of varied form and reflectance (Fig. 4). The residual should lie largely in the Munsell notations assigned to the individual samples, and it is the smoothing out of these individual sample errors which constitutes the principal task of the subcommittee.
The subcommittee did not undertake to evolve a single scaling unit for the entire solid, or for any attributive dimension of it. As is well known, the Munsell scaling units for hue, value, and chroma are far from equivalent perceptually. The relation between these units is, and remains, roughly as follows: 1 value unit=2 chroma units=3 hue units (at /5 chroma) (59). Even were the three units equated at the /5 chroma