Scientific Visualization has proven to be an effective means for analyzing multivariate multidimensional data (MVMD). A variety of techniques combining statistical and visual analytic tools have been developed in the recent years to analyze MVMD. Visual differencing, or visual discrimination, is the ability to compare an attribute value between two or more objects in a visualization. In this research, we are examining humans’ predictable bias in interpreting visual-spatial information for comparison and inference. We will develop and evaluate new techniques of data representation that support multivariate multidimensional visual differencing. We will also address the trade-off between proximity and occlusion and evaluate users’ ability to explore MVMD across the immersive spectrum.