We apply multivariate error-propagation analysis to color-signal transformations. Results are given that indicate how linear, matrix, and nonlinear transformations influence the mean, variance, and covariance of color-measurements and color-images. Since many signal processing paths include these steps, the analysis is applicable to color-measurement and imaging systems. Expressions are given that allow image noise or error propagation for a spectrophotometer, colorimeter, or digital camera. In a computed example, error statistics are propagated from tristimulus values to CIELAB coordinates. The resulting signal covariance is interpreted in terms of CIELAB error ellipsoids and the mean value of color-difference measures, Delta E-ab* and Delta E-94*. The application of this analysis to system design is also illustrated by relating a Delta E-94* tolerance to equivalent tristimulus-value error statistics.