Color appearance models allow for the quantification of color appearance under a variety of viewing conditions. Such models may ultimately provide a measure for accurate assessments of the color-rendering properties of light sources. This article evaluates the Hunt94 color appearance model using a new set of color-naming and magnitude-rating data. At one photopic level (10 cd/m(2)), the evaluations showed that for a xenon lamp and an enhanced metal halide lamp that have chromaticities and spectra close to an equal energy spectrum, the Hunt94 model provided good predictions of the primary and secondary color names and hue magnitudes for a wide range of color chips tinder the two illuminants. However for other light sources the Hunt94 model predictions deviated considerably from the evaluations. Three modifications were applied to the Hunt94 color appearance model to predict color-naming and magnitude-rating data better for all light sources. The modified Hunt94 model gave good predictions (correlation coefficients r similar to 0.95) of the secondary hue magnitude of the color chips used in the experiment at photopic light levels (10 cd/m(2) and 1 cd/m(2) background luminances) under "white" light sources. However, the modified model was still unable to predict color appearances at low mesopic light levels (0.1 cd/m(2) and 0.01 cd/m(2)). (C) 2005 Wiley Periodicals, Inc. Col Res Appl, 30, 107-117, 2005; Published online in Wiley InterScience (www.interscience.wiley.com).