Evolutionary models and absolute ranks
The criteria that have been, or could be, invoked to justify and determine absolute ranks include the geological age of origin, the number of included lower‑ ranking taxa (or of species), and phenotypic distinctiveness. Simple examples should convincingly demonstrate that none of these criteria has ever been, nor ever could be, widely applied to determine absolute ranks.
Number of included terminal taxanext section
The number of included terminal taxa (usually called ‘species’) or of lower‑ranking taxa could be used to determine ranks if the tree of life were completely symmetrical. In that case, we could adopt conventions to regulate the rank based on this criterion, possibly based on the log2 of the number of included terminal taxa (if we assume that cladogeneses systematically yield two rather than three or more daughter lineages). For instance, the number of species included could be 2 for species groups, 4 for subgenera, 8 for genera, 16 for subtribes, etc. (Fig. 2a). This could hold even if extinction subsequently eliminated some of these lineages. This convention might be workable if we had exhaustive knowledge of the tree of life (needless to say, this is far from being the case, except for extant members of some of the most intensively studied taxa). Unfortunately, nature does not produce symmetrical trees (Fig. 2b; Purvis and Agapow, 2002). It has been shown that the number of included genera in families is highly variable, ranging from one to more than 400 (Bertrand et al., 2006: fig. 2), and the same phenomenon prevails at other taxonomic levels, such as the number of genera per order, or the number of species per genus (Dial and Marzluff, 1989). Despite efforts in this direction (Van Valen, 1973), the number of included taxa of a given rank cannot be used to rank more inclusive taxa.
Geological age of origin
Perhaps the most frequently proposed criterion to rank taxa generally (not at a single taxonomic level) is geological age of origin of clades, and as such, it deserves a more detailed discussion. It was first proposed by Hennig (1966), and supported most recently by Avise and Mitchell (2007). Under that proposal, the geological age of origin of the least inclusive node subtending all terminal taxa of a clade (or even possibly paraphyletic group) would determine the rank of that taxon. For this purpose, geological times would be subdivided into a number of periods that could (but need not) coincide with currently recognized geological periods. The tremendous advantage of this proposal is that taxa of a given rank would have comparable ages of origin, a feature that would vastly improve usefulness of biological classifications in evolutionary studies, as emphasized by Avise and Johns (1999). The main drawback of this proposal is that it would require numerous changes in rank allocations, and this probably explains why this proposal has not been more generally followed, although some studies applied its principles to suggest rank allocation of a few taxa (Lim, 2007; Tinn and Oakley, 2008). The more recent proposal by Avise and Mitchell (2007) is aimed at avoiding this problem; it consists in appending timeclips to taxa (ranked or not). These could take the form of a three-letter code within brackets, such as [G:pa] to denote a clade originating in the Paleocene (the particular temporal divisions adopted and corresponding codes were left open by the authors). Thus, as this temporal information becomes available through molecular (e.g. Hugall et al., 2007) or paleontological dating (e.g. Maranović and Laurin, 2007), it could be incorporated into taxonomies. Under this latest proposal, rank allocation would not need to be changed to reflect the geological age of origin (although it could be done if the systematic community chose to do it), so it would not create taxonomic confusion. However, unless categories were reassigned to reflect age of origin, they would remain ontologically empty. The useful ranking information would reside entirely in the timeclips, rather than in the Linnaean categories.
The geological age of origin would be optimal to determine ranks if taxa originated in a synchronous, periodic (but not necessarily regular) manner. Suppose, for instance, that the first cladogenesis in a lineage occurs at time t. The next cladogenesis in both descendents occurs at time t+c1 (c1 is a constant). The four resulting lineages speciate again at time t+c2 (c2 is another constant). In such a case (Fig. 2a), since several taxa appear simultaneously, their age of origin could be used to determine their rank. Unfortunately, cladogeneses do not appear to be coordinated in such ways (Fig. 2b). Nothing in modern evolutionary theory (Lee and Doughty, 2003; Minelli, 2007; Padian, 2008) predicts that cladogeneses should be simultaneous. The age of origin of taxa is often difficult to determine, but both molecular (Sanderson, 2002; Hedges and Kumar, 2009; Hugall et al., 2007) and paleontological dating (Marjanović and Laurin, 2007, 2008) suggest that asynchronous cladogeneses are the rule. There are periods of intense cladogenesis, for instance when taxa invade new niches (Ward et al., 2006), or after mass extinction has emptied ecological niches (Bromham, 2003), but these presumably represent periods of dense, asynchronous cladogenetic events.
Ranks traditionally attributed to taxa certainly do not reflect geological age, as shown by even a cursory glance at the literature. Extinct organisms of any geological age are usually attributed to taxa of all five ‘mandatory’ categories (genus, family, order, class, and phylum), in addition to a taxon of species rank (Laurin, 2005). Thus, the oldest species is as old as life itself, and so is the oldest genus, family, order, class, and phylum. Even if we exclude extinct organisms (a decision that would be difficult to justify but that might somewhat improve the correlation between taxonomic rank and geological age of origin), the geological age of taxa of any given rank is highly variable. In these comparisons, the age of origin of a taxon will be taken as the age of its basal node (i.e. the age of its oldest fossil member, or the nodal age inferred by molecular dating), rather than the age of its stem, although changing this choice would only make all taxa older without changing the age difference between them much. Sirenidae (ranked as a family), a clade of aquatic salamanders, originated in the Early or Late Cretaceous (about 80 to 110 Ma ago), depending on whether or not some extinct forms are included (Marjanović and Laurin, 2007: fig. 3). Hominidae (also a family under rank‑ based nomenclature) originated about 7 Ma ago in the Miocene, if it is defined as the largest clade that includes Homo sapiens but not Pan troglodytes (Linnaeus, 1758), the chimpanzee (Pilbeam and Young, 2004), so it is at least 10 times more recent. Some lissamphibian genera (Amphiuma, Necturus, Dicamptodon) appeared in the Paleocene (Marjanović and Laurin, 2007: fig. 4), about 60 Ma ago, whereas the genus Homo dates from less than 3 Ma (Semaw et al., 2005). Clearly, the geological age of origin of taxa of a given rank is highly variable; changing rank allocation to improve the correlation between age and rank would result in so many, and so drastic, nomenclatural changes that this solution will surely appear unsatisfactory to most systematists.
Finally, phenotypic distinctiveness could be used to determine absolute ranks objectively (Mayr and Ashlock, 1991) if nature proceeded by discrete steps, or by gradual evolution under special circumstances (for instance, if cladogeneses were synchronous and if evolution proceeded at a steady rate). For instance, under a speciational or punctuated model of evolution, if the amount of phenotypic change could take only a few discrete values (not necessarily multiples of each other), phenotypic change could be used to assess the absolute rank of daughter lineages, at least for a small set of taxa (Fig. 2a). However, neither evolutionary theory nor observations corroborate any such evolutionary model; instead, the magnitude of phenotypic gaps appear to be highly variable (Fig. 2b). Even when the evolutionary model appears to be speciational or punctuated (Cubo, 2003; Mattila and Bokma, 2008), there is no evidence that the amount of phenotypic (or even genotypic) change takes a limited number of values, and of course, in most cases, gradual evolution presumably plays an important role, instead of, or in addition to, speciational change. Thus, phenotypic distinctiveness cannot be used to assess absolute ranks. It is difficult to show that phenotypic distinctiveness is highly variable between taxa, because it is difficult to quantify, but the very fact that it has not been quantified for most taxa (see Wills et al., 1994, for some exceptions) suggests that it has not been used as a criterion to rank taxa, or if used, only very imprecisely so.
If any regularity in the evolutionary model prevailed (if cladogeneses were synchronous, or if the tree were symmetrical, or if phenotypic gaps were discrete), ranks could perhaps be assigned objectively (Fig. 2c, d). However, as the above review shows, evolutionary theory and observations fail to confirm any of these special models. This leaves us without general rules to potentially assign ranks objectively.