A review of some literature concerning traffic sign illumination is given. It is concluded that there is good agreement between various studies of luminance and legibility, but that empirical data concerning the relation between luminance uniformity and legibility are completely lacking. As a result an experiment was made to investigate this relation. The results indicate that there is no correlation between luminance uniformity and legibility on suboptimal luminance levels, but that there exists a significant correlation on the optimal luminance level. From this relation some recommendations concerning acceptable luminance uniformity (expressed as luminance ratios) are derived.