The affect of purity on hue has been reported severally (including Munsell and NCS data on constant hue loci) over 100 years but without general agreement. For example, the number of hue shift nulls in the data vary from 2 to 6. Hence, despite this effect's commonality it lacks reliable data for modeling underlying mechanisms or color appearance. The purity of a stimulus may be decreased by adding white (as in Abney's experiment), by adding black, or by adding gray such that luminance, or alternatively lightness, remains constant. This article gives new data for CRT stimuli for illuminant D65 for all four conditions but mainly for equal luminance, for 31 observers and 13 test dominant wavelengths. Further, samples were observed in two temporal conditions: either simultaneously as pairs (the contrast mode) or singly (the no-contrast mode). Three types of samples were tested: (1) equal luminance 30 cd/m(2) for all dominant wavelengths, (2) equal lightness for all dominant wavelengths, and (3) zero-gray colors, requiring dominant luminances for different dominant wavelengths. In all the above conditions, the resultant hue shifts graphed a robust bimodal curve (two peaks in cyan and red, two troughs in blue and green) across the hue cycle, similar to Munsell and NCS data except the definite peak in cyan. (c) 2006 Wiley Periodicals, Inc.