As the several specific applications in this paper demonstrate, multidimensional scaling provides a long-needed means for investigating and describing spatial relationships among speech varieties. It is especially applicable to the relationships among varieties of a single language (or more properly, linguistic "cline"), which, as is generally known, are poorly described by the hierarchical mode of classification commonly used in comparative linguistics. But multidimensional scaling may also frequently be used to describe spatial variation which has persisted among distinct but related languages and which cannot be adequately described by an otherwise well-motivated hierarchical classification. These conclusions are illustrated by the application of multidimensional scaling to lexicostatistical percentages within four linguistic groups, located in the Philippines, Africa, and North America. (Author)