Ensuring perceptual conditions that the driver is to be able to detect a critical object ahead, at a sufficient distance, to take necessary driving manoeuvres safely, an attempt was made to find the admissible limit of the uniformity of the road surface luminance (U=Lmin/Lr) in relation to the average road surface luminance (Lr), under full scale road lighting installations with actual driving. (Lmin is the luminance of the darkest part of the road surface.) In the experiments, in total, 56 different combinations of U and Lr were provided and observed. Applying a forced choice method, correct responses of the observers in the car driven at a constant speed of 60 km/h, concerning the lateral positions of object(s) located against the darkest part of the road surface, presented in a random ordar, were collected and analyzed for each combination of U and Lr. Based on the data collected, probabilities of the correct perception of (lateral position of) the critical object were derived. Finally, the admissible limit of U is given as a function of Lr, to perceive the object with a probability of 75%, taking the luminance contrast of the object as a parameter. Consequently, it has been found that the square of the admissible limit of uniformity varies in proportion to a ratio between Lu and Lr, i.e. U2=Lu/Lr, where Lu is the luminance of the road surface necessary to perceive the object when U=1.0. The relation given agrees well with that obtained with the lighting simulator, and those derived from the results of previous researchers.