One of the questions most commonly asked when looking at a phylogenetic tree is what that tree indicates about how the organisms on it evolved. What does it say about what their ancestors were like? What changes happened when? In answering those questions, the most commonly invoked tool is the principle of parsimony - in the absence of any reason to think otherwise, favour the explanation that is the most straightforward, and (in the case of inferring evolutionary history) requires the least number of changes. Parsimony is a popular tool because it's straightforward, relatively easy to apply, and it makes a great deal of intuitive sense - if a red animal occupies a deeply nested position in a clade of blue animals, then it seems fairly obvious that the ancestral animal was blue. However, like all analytical tools, the principle of parsimony is based on certain assumptions, and can be misleading if those assumptions are violated. Parsimony assumes that when comparing changes in a character between two states, change in either direction is equally likely. If, for whatever reason, a change is more likely to happen in one direction than another, then a parsimony analysis might be mislead about the ancestral condition. An elegant demonstration of this limitation of parsimony can be found in Collins et al. (1994) were the results were discussed of using parsimony to infer an ancestral DNA sequence (cytochrome b) for the marine gastropod genus Nucella. Nucella DNA is AT-rich (its base composition includes far more As and Ts than Gs and Cs). Inferring the ancestral sequence using parsimony implies an ancestor even more AT-rich than any of its descendants, despite the fact that the AT-bias remains fairly constant across all living members of the clade. Because GC bases are relatively uncommon for a given position, the parsimony analysis always tends to indicate them to be the derived state.
A few years ago, a paper appeared in Nature presenting a phylogenetic analysis of Phasmatida, stick insects or phasmids (Whiting et al., 2003). Phasmids include both winged and wingless taxa. In those taxa that do have wings, the forewings are greatly reduced and the hindwings are the functional pair. Whiting et al. (2003) found that the various winged phasmids were phylogenetically nested well within a series of wingless taxa. They therefore made the surprising suggestion that the common ancestor of living phasmids was wingless, and that those phasmids with wings had regained them secondarily.
As remarkable as this may sound, it may not be impossible. Studies of embryonic development in animals have found that developmental regulatory genes often act in a hierarchical manner, so that a relatively small mutation in a gene acting on an early stage of development may have a significant effect on later stages of development. It has therefore been suggested that it might be possible for a given feature (such as wings in an insect) to be lost through a mutation causing that feature to start developing in the first place, without any change in the genes that shape how that feature develops once it starts. If the original regulatory gene was then to mutate back to its original condition in a later generation, the missing organ might spring back into its original position as if it had never left. It might be argued that the patterning genes rendered functionless by the original mutation might be suject to genetic drift, and degrade to useless pseudogenes that could not be reactivated even if the original mutation did revert, but Whiting et al. (2003) invoke another of the interesting features of developmental genetics - many genes are pleiotropic (involved in patterning different characters at once). For instance, insects use many of the same genes in patterning their legs as patterning their wings, so even if selective pressure to retain function for the one was removed, there still might be the need to retain function for the other.
While this may be theoretically possible in general, is it the case for phasmid wings in particular? Does the evidence offer strong support for regained wings in stick insects? Despite Whiting et al. (2003) being widely cited as a proven case of evolutionary regain of a complex character, I'm going to have to answer with a no, I don't think so. Stick insects are generally not highly mobile. Even those species that have fully functional wings fly only rarely. They are exactly the type of insect that one would expect to be prone to frequent flightlessness and wing loss. Whiting et al. (2003) themselves state at one point that a winged ancestor for crown phasmids became the most parsimonious reconstruction if wing loss was weighted as six times more likely than wing gain. This does not seem too unlikely a difference. Other potential evidence can be seen in the wings themselves. Whiting et al. argue that genes involved in wing development may have remained potentially functional if they were still being used for other organs. But how far can this argument be taken?
One of the most useful features in characterising insect wings is their venation. A generalised diagram of insect wing venation is given above, but different orders of insects have significantly different wing venation, enough so that relationships can be recognised for fossil insects known from wings alone. Comparisons between wing venation of different orders can also be very useful in establishing their relationships.
Wing venation of phasmids shares a number of distinctive characters with that of Orthoptera (grasshoppers and crickets). Both these orders have the forewings leathery, with the main veins running roughly parallel. It is the hindwings, however, that show the major similarities (Grimaldi & Engel, 2005). In both orders, the cubital veins (the veins marked Cu and in blue in the diagram above) don't run to a point low on the hind edge of the wing as they do in most orders, but instead run fairly straight out to near the distalmost tip of the hindwing. The veins in front of the cubitals, which enclose most of the wing space in other insect orders, are packed into the fairly small space between the cubitals and the front of the wings (this is the hardened part of the wing in the photo at the top of this post). Most of the hindwing in orthopterans and phasmids is composed of the anal fan, reasonably small in other insects but massively expanded in these orders. All veins in both wings are densely connected by numerous crossveins. The close relationship between phasmids and orthopterans suggested by these shared characters has also been supported by molecular analyses (Terry & Whiting, 2005), albeit with the inclusion of the webspinners, which have greatly simplified wings with a much-reduced venation. While pleiotropy might explain how wing-patterning genes remained functional overall, it is difficult to imagine how the form of the potential wings could have been maintained down to their very venation.
For contrast, Alex Wild recently discussed a much better-supported case of character reversal. Army ants of the genus Eciton have functional eyes in the workers despite being descended from an eyeless ancestor. However, while other ants have eyes with well-marked facets and multiple ommatidia (lenses) like those of other insects, Eciton eyes are nowhere near as well organised. The separate ommatidia have become atrophied and fused together, so that unless examined at electron microscopic level they look like a single enlarged ommatidium. Eciton worker eyes resemble the eyes of other insects the way that a six-year-old child's drawing of a horse looks like a real horse. You can see that the idea's there, but the execution is still something of a shapeless blob. What makes this situation even more remarkable is that there can be no doubt that Eciton still possesses the genes for growing fully-formed eyes, because the winged males (which never lost their eyes in the first place) still have perfectly normal insect eyes.
While pleiotropic selection might preserve the overall position and maybe even shape of the wings, there seems little reason for it to preserve the fine detail. After all, there are countless different ways that wing veins could potentially be arranged to give similar shape and function - that's how venation can vary so much between orders in the first place. Even if loss and regain of wings in phasmids might seem the most parsimonious explanation, I just don't think that it is more convincing than the alternative suggestion that phasmids have a repeated bias towards wing loss.
Afterword: I had written all this before I found the commentary on Whiting et al. (2003) by Trueman et al. (2004), and the reply by Whiting & Whiting (2004). I'd recommend reading them.
Collins, T. M., P. H. Wimberger & G. J. P. Naylor. 1994. Compositional bias, character-state bias, and character-state reconstruction using parsimony. Systematic Biology 43 (4): 482-496.
Grimaldi, D., & M. S. Engel. 2005. Evolution of the Insects. Cambridge University Press: New York.
Terry, M. D., & M. F. Whiting. 2005. Mantophasmatodea and phylogeny of the lower neopterous insects. Cladistics 21: 240-257.
Whiting, M. F., S. Bradler & T. Maxwell. 2003. Loss and recovery of wings in stick insects. Nature 421: 264-267.