Principal components analysis is often used to fit a population of spectral reflectances by a mean vector plus a basis-function expansion about the mean. Certain color-technology applications (such as color correction) are much easier if the mean is absent. If the mean of reflectance (or of another spectral function) is a linear combination of the first few principal components (such as the first three), then a linear model can fit the original data without mentioning the mean vector in the model's,formulation. This idea is worked out step by step, and a realistic example is presented. (C) 2004 Wiley Periodicals, Inc.