Three experiments were performed to evaluate methods for predicting the luminances and chromaticity coordinates produced by color CRT monitors, given known inputs. Linear and logarithmic versions of PLCC and PLVC, plus Ber ns, Motta, and Gorzynski＇s power function, were tested. Estimates are provided for the number of CRT,measurement points needed to maximize each method＇s predictive accuracy. Correcting for dr,unintended light from the monitor is shown to improve accuracy substantially for a case involving a seemingly small amount of light. Berns et al.＇s characterization technique, which involves measuring the,monitor＇s neutral point, is shown to yield the same accuracy as conventional characterization while reducing the number of measurements required, and to yield improved accuracy when correction for unintended light is needed but impractical. The accuracies of the predictive method＇s are compared and recommendations for their rise are provided, (C) 2000 John Wiley & Sons, Inc.*