Two fundamental questions relating to a proposed engineering approach for a color reproduction system are discussed. These questions are: (1) What is the mathematical relationship between how a color original looks and the amounts of the printing colors required for a visual match?; and (2) How much variability would be observed in a controlled printing operation if a set of plates was printed from time to time over a fairly extended period of time? The first question is answered in the form of an empirically determined second degree transformation equation. In effect, this equation relates a transformed CIE color vision space to an E.N.D., amount of colorant, space. Variability data was obtained from a controlled color printing operation over a 3-month period of time. The transformation equation was then tested by using it to predict the amounts of colorant required to match original paint swatches. Plates were made to print these predicted amounts. The resulting reproductions were compared to the originals and found to match quite closely. The observed differences are in the order of magnitude of the tolerances used for commercial trademark colors. Variability data was used in the inverse transformation equation to calculate CIE Limit Solids. These CIE Limit Solids represent the #3 standard deviation limits beyond which the reproduced color is not expected to lie. All eleven reproductions fell within these CIE Limit Solids. It is concluded that an engineering approach to color reproduction is quite possible and that such a system could be built with existing technology. Reproductions could be made from reproducible copy with an accuracy limited mainly by the press induced variability.