WHAT CAN FOSSILS REALLY TELL US ABOUT PHYLOGENY?

Some estimates suggest that there are fifty times as many extinct species as extant ones. Many large branches of the tree of life are known only from fossils, and the data that they preserve may therefore be essential for properly understanding phylogeny, evolutionary patterns and transitions. However, the notorious incompleteness of fossils - along with the greater need to interpret extinct morphologies - mean that they are often suspected of providing data of lower quality, or of introducing more homoplastic noise into analyses than extant species. While we find no evidence for this in general terms, the preferential loss of soft-part characters prior to fossilization may result in 'stemward slippage': the process by which partially decayed specimens erroneously resolve closer to hypothetical ancestors or outgroups.

We also find that paleontologists preferentially focus on particular anatomical regions; sometimes for practical reasons and sometimes believing them to convey a stronger phylogenetic signal. In particular, vertebrate systematists often concentrate on the skull at the expense of the rest of the skeleton. We observe no differences in levels of homoplasy between these partitions in published data sets, but find that the optimal trees derived from them are significantly different more often than we would expect; a result that may indicate differing patterns of homoplasy and convergence across body regions. This suggests that analyses should sample characters as widely as is possible.

Finally, since fossils convey information on temporal (stratigraphic) distributions as well as on morphology, trees are often calibrated against time in order to elucidate patterns and rates of diversification and extinction. The congruence between phylogenetic branching order and stratigraphic first occurrence order is significantly different across higher taxa. Poor congruence may reflect inaccurate trees, a patchy record or both, and we consider the utility of indices and tests intended to highlight such issues. Despite its enormous utility, therefore, we conclude that fossil data - like all systematic data - must be evaluated with some caution.