In a series of experiments observers judged whether two visual tachistoscopically presented shapes were the same or different in a relevant dimension, and had to ignore the graded variation on an irrelevant dimension that appeared concurrently with the relevant dimension. Experimental results from judgments in orientation, size and brightness failed to support the normalization hypothesis. The hypothesis predicts a monotonous increase in RTs with the increasing degree of disparity in the irrelevant dimension in same as well as in different comparisons. The results were interpreted in terms of the type of dimensions used to construct the shapes. It was suggested that with separable stimulus dimensions normalization would not be necessary. However, interference might appear when the stimuli to be compared were generated from a combination of more integral dimensions.