To assist genome annotators in naming transposable elements (TEs), Wicker et al. propose in their Guidelines article (A unified classification system for eukaryotic transposable elements. Nature Rev. Genet. 8, 973–982 (2007))1 a classification and nomenclatural system aimed at simplifying annotation to assist future detailed structural, functional and evolutionary analyses1. However, even though the proposed classification system mimics organismal phylogenetic classification, it fails to do so consistently.

Classifications are a necessary prerequisite for human communication, and most biological classifications are hierarchical. The relationships between entities in any hierarchical system are of two kinds only: either one entity is included in another or they are mutually exclusive. The only rule that applies to the ranks in biological nomenclature is their sequence — a class includes one or more orders, an order one or more families and so on. Most biologists agree that classifications should reflect phylogeny (see Ref. 2 for one exception), although how phylogeny is reconstructed is controversial. This aside, the majority of organismal phylogenies are inherently hierarchical, and thus either a clade is included in another clade or two clades are mutually exclusive. Hence, as long as the groups remain monophyletic3 (for an alternative point of view see Refs 4, 5), a hierarchical classification is ideally suited to reflect phylogeny6.

Some groups with formal rank in the TE classification proposed by Wicker et al.1 seem to be well justified (for example, classes). However, the limits of groups become fuzzier as one descends through the hierarchy and, to an increasing degree, loses connection with phylogeny. Some groups are surely not monophyletic; for example, autonomous and non-autonomous partners of the same family are classified as different subfamilies. In addition, although recognizing non-monophyly of miniature inverted repeat elements (MITEs), Wicker et al.1 repeat the classification error by recognition of paraphyletic subfamilies that include only autonomous members, even though they claim that “subfamilies are defined on the basis of phylogenetic data”1. Furthermore, the 80–80–80 similarity rule that is used to create families and subfamilies leads to an unpredictable mix of monophyletic, paraphyletic and polyphyletic groups. Naming paraphyletic groups may be misleading3, but naming polyphyletic groups has no connection to reality.

Additionally, it must be remembered that, despite their mobility, TEs over time 'travel' within their host genome. Hence, although a TE phylogeny corresponds to a gene tree and need not reflect species phylogeny, the evolutionary histories of the TEs and of their hosts are tightly linked, and will need to be integrated to achieve a full understanding of both7.

It is particularly discouraging that Wicker et al.1, despite being aware of the fact that homology is explained by common descent, misuse8,9 the concept by reference to “a continuum of sequence homology”1. Homology (as with pregnancy, for example) is not measured in degrees: either two sequences are homologous or not. However, in contrast to pregnancy, homology is only a hypothesis8,10,11,12.

We absolutely accept the need for a classification system of TEs, but equally urge that such a system must be based on the consistent use of principles. Anything else may retard, rather than improve, our understanding of TE evolution.