This study is aimed at showing that the fractal geometry of taxonomic systems (Burlando, 1990) reflects self-similar evolutionary pattern. Evidence is achieved by three steps: (i) examination of taxonomic data from the fossil record; (ii) examination of taxonomic data from phylogenetic systematics; (iii) comparisons among different levels of the taxonomic hierarchy.
In each step, all or nearly all the examined assemblages yield frequency distributions of numbers of subtaxa within taxa which fit a hyperbolic model function, confirming the fractal pattern. The first two steps show that the pattern is not deriving from classification bias, while the third one verifies the self-similarity of evolutionary radiations. According to the first and third step, self-similar cladogenesis consists in the arising of many isolated lineages and clumps of lines, the latter consisting of isolated lines and clumps, and so on. The properties of fractals led to the hypothesis that scaling diversity emerging from taxonomy could actually encompass the species level, thus limiting the importance of species within the evolutionary context in favour of a more comprehensive view of life diversification.