Looking at more than three dimensions

Statistical graphics for data in more than three dimensions cannot be based on direct spatial representation. Three approaches have been tried with varying degrees of success. First is simultaneous presentation of two- or three-dimensional marginals with a cross-reference scheme to identify common points. Second is projection into two or three dimensions and variations on this theme. Third is a single view with higher dimensions indicated by characteristics of the plotting symbol. The first two approaches admit visualization of simple conditional relationships, but putting together pieces to understand complex relationships requires substantial intellectual effort. The third is a compromise which presents some visualization of higher dimensional relationships. The paper emphasizes the third mode of display making use of motion, color, and geometrical characteristics of the plotting symbol as higher dimensional cues. 19 references, 4 figures.