As good singers know, an important early step is learning to hear. Similarly learning to see is important in designing graphics. When one looks at scientific summaries through the eyes of human cognitive limitations, graphical design puzzles abound How can graphics be optimized to communicate scientific patterns to human beings? Graphical design principles provide guidance that lead to interesting places if one drops the principle that says: follow convention. Following convention is helpful to cognition, but today's innovation can be tomorrow's convention. The puzzle solver should be free to too use all tools including a supercomputer to find human optimal layouts and revealing views. This talk calls attention to several graphical design challenges and proposes some solutions. One set of examples concerns the study of gene expression clusters and representation of gene header sequences. Another set concerns the representation of multivariate environmental summaries for Omernik's ecoregions. A third topic concerns the redesign of Trellis( graphics strip labels. The talk concerns both work in progress and new graphics, such as linked micromap bivariate boxplots and Portuguese man-of-war boxplot glyphs, that are ready to be put to work in other applications.