Page:Journal of the Optical Society of America, volume 33, number 7.pdf/62

 an admixture of a slight degree of rod vision. The Munsell samples, on the other hand, have been observed under a degree of light adaptation insuring a greater approach to pure cone vision, but with fields so greatly exceeding two degrees that the macular pigment probably exerted little influence. Perhaps these differences in observing conditions account for the peculiarity in the purple-blue region. Since definite justification for a sudden transition is lacking, the curves were not altered as radically as they might otherwise have been.

Originally it was believed that the chromatic-value charts would also simplify the smoothing of the constant-hue lines by permitting them to be drawn with a straightedge. The number of irregularities to be contended with, however, soon made it apparent that the simplest thing to do was to smooth directly in the I.C.I. (x, y) diagram. Accordingly, the visual estimates for the 20 principal hues, corresponding to the 1929 Munsell samples, were plotted in vector arrow notation with respect to both the National Bureau of Standards and the Glenn-Killian points; and then tentative loci of constant hue were drawn in for each value level.

An attempt was made in the first smoothing to hold to lines of constant dominant wave-length, but degrees of curvature which could not be ignored were soon apparent in some hue lines. When the curved smoothed lines of hue for separate value levels were put together, a picture of fairly regular progression emerged (Fig. 13). In numerous instances it was necessary to extrapolate the hue lines far beyond the Munsell data to reach the theoretical pigment limit. One working rule was to carry to the limit the trend of the curve which seemed best to fit the data. Another rule was to err in the direction of straightness rather than of curvature. Of course, the loci for a given hue on different value levels had to be smoothed and adjusted to provide smooth inter-value-level transitions. Tracings of the several loci for a given hue were compared and adjusted, after which the smoother reviewed the arrangement at each value level. Thus, as in the chroma smoothing, by reverting back and forth from the intra-value to the inter-value operation, discrepancies were progressively reduced until the optimal hue lines were eventually approximated.

In certain extended regions beyond the Munsell data, the problem of determining the exact courses of the hue loci is formidable. In some instances, fortunately, the results of other studies were available to supplement the visual estimates, and proved helpful in reaching or confirming a decision. In spite of our practical precept favoring straightness, observations of strong colors by Judd (17), and by Kelly and Judd (20) made it clear that a number of our provisional curved loci should be increased in curvature toward their extremes. In particular, the PB line was kept nearly straight while the 7.5PB line was altered from nearly straight to considerable curvature toward its limit on the basis of Judd’s previous data and check observations by Judd, Nickerson, and Newhall. The greatest curvature occurs immediately following the PB, and is not shown since it lies between the PB and 7.5PB lines. Hue lines between 7.5PB and 5.0P were adjusted toward their ends to conform with a curvature gradient decreasing toward P, as observed by Kelly and Judd. Their observations are also responsible for increased curvature of the line ending at 620 my on the spectrum locus. This line represents a local curvature maximum, the loci on either side being altered slightly to conform to a progressive change in curve.

There is also older evidence that the loci of constant hue cannot be expected to coincide with those of dominant wave-length, even when luminance is constant. This question has been considered by Müller (33), Abney (1), Schrödinger (34), and Judd (18). Hue appears to be an imperfectly known function of purity when luminance and dominant wave-length are constant. Around the spectrum locus the effect seems to be in different directions in different regions, with several hues invariable, or nearly invariable in this regard. In Fig. 13 an invariable (straight line) at 10Y and another between 5P and 7.5P may be noted.

The recommended hue loci also seem to be affected by the well-established Bezold-Brücke phenomenon (5), (6), (14), (35), (32), that is, the hue shift with luminance, apart from purity and