For millennia, taxonomists have tried to make sense out of the world by visually and verbally organizing it. In biological terms, pre-evolutionary thinkers have often suggested an affinity of humans to other mammals, among other relationships expressed in their taxonomies. The availability of a well-corroborated tree for living mammals enables me in this chapter to ask if past authors succeeded (or not) in creating accurate taxonomies, and if specific methodological and theoretical advances (e.g., the recognition of evolution via descent with modification) are associated with improved taxonomies. They are. I find that different methods used over past centuries for building taxonomies (including pre-evolutionary, evolutionary, cladistic, and molecular) improves on its predecessor, although at varying rates and levels of significance. H.M.D. DeBlainville was an outlier among pre-evolutionary authors and produced substantially more accurate taxonomies than his contemporaries. Evolutionary authors constructed significantly more accurate taxonomies than their predecessors, and began at a higher level of accuracy than would be expected based on previously observed improvement over time. Cladistic methods show improvement at a greater rate than previous methods, but do not start at a level greater than expected based on improvement evident among pre-cladistic authors. Authors using molecular data, including datasets independent of the nuclear DNA key for recognizing the currently well-corroborated tree, began with levels of accuracy similar to those applying cladistic methods to morphology, but thereafter showed the greatest improvement in the shortest amount of time. All evolutionary methods (including cladistic and molecular) significantly outperform pre-evolutionary methods in recognizing well-corroborated mammalian groups.