The phylogeny of Odonata cannot be estimated separately from the evolution of the odonate wing. A variety of wing-vein naming systems have been proposed during the past 100 or so years. Each implies its separate view of odonate higher relationships. Likewise, a variety of phylogenetic hypotheses have been proposed, and each of these implies a view of historical developments in the venation.
The Comstock and Needham (1898) wing vein naming system, still widely used by dragonfly workers in the United States, implies that the common ancestor of modern Odonata was anisopteran and that the zygopteran venation is arrived at by reduction. The Tillyard (1917) version of this system implied Zygoptera is sister to Anisoptera but reconciled wing vein patterns across the suborders by giving different names to veins in the same position on the wing. That modified version is not in current use but has influenced some past estimates of the phylogeny.
The quite different Tillyard (1926, 1928, 1935) system, still much used by dragonfly workers outside of North America, is based directly upon a phylogenetic hypothesis that Protozygoptera and Zygoptera are grades on the way to Anisozygoptera, and that Anisozygoptera is a grade on the way to Anisoptera.
A few authors (eg Lemsche, 1940; La Greca, 1980; Matsuda, 1981) have suggested that the wings and hence wing veins of Odonata are not homologous with those of other insects. This conclusion appears untenable, although, to judge by the variety of flight muscle mechanisms known in insects, flight may have evolved more than once.
Forbes (1943) attempted to develop a common venational notation for all insects. His scheme as modified by Hamilton (1972) homologises the discoidal cell in Zygoptera with the triangle in Anisoptera. In contrast, all other major systems homologise the triangle plus supertriangle with the discoidal cell. Martynov (1930) regarded the vein that closes the anisopteran triangle as derived convergently in different anisopteran families, which would make Anisoptera polyphyletic (although according to Martynov the extant families are closely related and form a natural group).
New systems of vein homology were developed by Carle (1982), and Riek and Kukalova-Peck (1984). Each of these systems is based on an assumption that the zygopteran venation is derived by reduction, either from one or more anisozygopteran ancestors (Carle) or from Anisoptera (Riek and Kukalova-Peck). The two systems each involve the hypothetical loss of some sections of main longitudinal veins, and secondary re-attachment of the distal parts of some veins to other stems. Trueman (1993) outlined a simpler system in which all the main veins known from other pterygote insects are present in Odonata and no vein is discontinuous. The Trueman system is set out in Gullan and Cranston (2000.p.38). The phylogenetic implication of that system, according to a cladistic parsimony analysis, is that Zygoptera and Anisozygoptera are paraphyletic and Anisoptera is the monophyletic sister group to some extinct Anisozygoptera. If this is correct (and it has yet to be tested) one implication for the history of wing venation would be that the expanded hind field of Anisoptera is secondarily derived and veins in that region of the wing are not homologous with the veins found in other insects.
The Riek and Kukalova-Peck (1984), and the Trueman (1993) systems each suppose that insect wings originated as primitively moveable appendages associated with precoxal leg segments. This idea or its precursors can be traced from Oken (1831) but finds its modern expression in Wigglesworth (1973, 1976), Kukalova-Peck (1978, 1983, 1985, 1987) and Trueman (1990). The other wing vein naming systems referred to above each assume an origin of wings from fixed paranota (Mueller, 1873-75; Crampton, 1916), an idea which may be past its time.
It is difficult to see how further progress in odonate higher phylogeny can be made, at least on the basis of morphological characters, without greater agreement on the identity of wing veins across the odonatoid orders and suborders. Perhaps for this reason, several research groups currently are turning to other character sets, eg, molecular sequence data, in hope of resolving this riddle.