Three approaches to the graphic representation of similarity and dissimilarity matrices are compared and contrasted. Specifically, Kruskal's multidimensional scaling, Johnson's hierarchical clustering, and Waern's graphing techniques are employed to depict, in two dimensions, data representing the structure of a set of botanical concepts. Each of the three representational techniques emphasizes a different aspect of structure in matrix data; specifically, content structure and cognitive structure are discussed. The researcher's assumptions regarding the structure of the data govern the choice of an appropriate representational method. When a hierarchical, class-inclusion structure is suspected, hierarchical clustering is the technique of choice. When underlying dimensionality is suspected or sought, and particularly when more than two dimensions are possible, multidimensional scaling is the proper technique. However, if there are no reasons for assuming the presence of a particular structure, Waern's graphing is the easiest and least expensive first choice. If a graphing solution yields promising results and two dimensions give adequate insight into structure, then combining the graph with a multidimensional scaling solution results in a representation with the greatest number of salient dimensions. (Author/MV)