There are practical reasons to represent a set Of reflectance spectra by a limited basis-function expansion. Such an expansion is most tractable in color technology applications when it is homogenous. However, more often a true principal-component analysis (PCA) is performed, which involves an offset of the expansion by a mean spectrum. In deference to custom, we discuss and exercise the traditional PCA on reflectances. We review the method of calculation and present tables of the principal components of 3,534 object-color samples from three color-order systems. Next, we describe two kinds of principal-component coordinates of a measured reflectance spectrum, one obtained by least-square best fit and the other by tristimulus match under one or more specified lights. Finally, we show illustrative examples, and propose some possible new uses for principal-component analysis of reflectances. (C) 2004 Wiley Periodicals, Inc.