A colorimetric printer model takes as its input a set of ink values and predicts the resulting printed color, as specified by reflectance or tristimulus values. The Neugebauer model has been widely used to predict the colorimetric response of halftone color printers. In this paper, techniques for optimizing the Neugebauer model are presented and compared. These include optimization of the Yule–Nielsen factor that accounts for light scattering in the paper, estimation of the dot area functions, and extension to a cellular model. A new technique is described for optimizing the Neugebauer primaries using weighted spectral regression. Experimental results are presented for xerographic printers using two halftone screens: the random or rotated dot, and the dot-on-dot screen. Use of the Yule–Nielsen factor, the cellular framework, and spectral regression considerably increase model accuracy.