Colors of objects are often specified by tristimulus colorimetry in some standard colorimetric system. While color-matching allows a color to be reproduced accurately, it does not specify what the color actually looks like. In fact, colorimetric matches will hold under widely varying conditions, while appearance may change drastically. Color-order systems, such as the Munsell, are often used for appearance descriptions but they suffer from several important limitations. A more direct approach is to use hue and saturation scaling, a technique which, when used correctly, has a precision not very different from that of color-matching.
The specific aims of this proposal are to refine the procedures of color scaling and to develop methods that will permit its use as a tool with which to study color vision. These scaling methods allow rapid collection of data, which makes it possible to obtain comprehensive findings across large parametric spaces. This is necessary for any detailed modelling of the visual system. The analytical methods will use multidimensional scaling to derive Uniform Appearance Diagrams from which standard discrimination functions may be calculated. These functions will be compared with functions measured by conventional methods. While these scaling methods are known to work with English, it is important to see if they are independent of the specific language used; thus they will also be tested on subjects whose native language is not English. The fully developed scaling techniques will be used to study the complete gamut of color space consisting of stimuli that vary in both wavelength and purity. Finally, the methods will be used for a detailed examination of peripheral color vision under a broad range of conditions. These experiments will provide a base of data about normal visual function and will provide the basis for determining which neural cell types underlie various aspects of color vision.
