Four data sets are analyzed to quantify three effects of luminance of samples on chromaticity discrimination: on ellipse area, axis dimensions (a and b), and a/b ratio. Ellipses for aperture, surface, and simulated surface colors in CIE 1931 and 1964 x, y, Y color spaces are shown to reduce axis dimensions with higher luminance by different functions for the major and minor axes. Reduction is greater, for major than minor axes, thus improving ellipse circularity. The,functions plot straight lines in log-log scale as power law equations, except luminances below 3 cd/m(2). We give formulae to predict a and b axes, a/b ratio, and ellipse area for almost any luminance in x, y, Y spaces. Effect of luminance is remarkable on ellipse area, which on average halves with every 3.5 times higher luminance. To illustrate the substantial effects of luminance, RIT-DuPont ellipses are predicted for three levels of equal luminance at 42, 212, and 2120 cd/m(2). In the latter, ellipses are much smaller and are nearer circular than in the former. Higher luminance is known to improve color discrimination, so reduced ellipse area is to be expected but does not occur in CIELAB and DIN99 spaces because of lack of luminance-level dependency. We discuss our results' implications on uniform color space. Weber fraction Delta Y/Y indicates brightness discrimination decreases with increasing luminance and is thus independent of chromaticity discrimination. (c) 2005 Wiley Periodicals. Inc.