Reconstructing the Tree of Life: Are molecular phylogenies really better than morphological ones?

The phenomenon of evolutionary convergence has intriguing implications for the extent to which
evolution may be predictable (https://theconversation.com/what-do-aliens-look-like-the-clue-is-inevolution-63899).
However, it often makes it difficult to reconstruct evolutionary relationships with
accuracy. In particular, it is often assumed that convergence is a problem that particularly bedevils
morphological phylogenies, making them less accurate
(http://www.42evolution.org/videos/researcher/professor-matthew-wills/), but is this really the
case?
It is increasingly easy to generate phylogenetic trees from molecular data, with analyses routinely
concatenating sequences from many tens or hundreds of genes. Despite their data-rich
underpinnings, new molecular trees often contain surprises, and frequently imply relationships that
are at odds with established and traditional hypotheses derived from morphological data. One such
example is the phylogeny of mammal orders: the 'new mammal phylogeny' overturns many
'traditional' groups (Fig. 1). Another example is the arthropods. How are we to view these conflicts?
Should we necessarily conclude that the new molecular trees are likely to be the correct ones; swayed
by the mass of molecular data and greater complexity of analytical models? To do so implies that many
decades of comparative anatomy and morphological scholarship have yielded data that are - at best
- simply more noisy and error prone than the data from molecules. At worst, it may indicate that
convergence within morphological data implies relationships that are positively misleading.
Since phylogeny cannot be known with certainty, there can be no objective test of accuracy. However,
it is possible to assess the congruence of competing trees with independent sources of data on
evolutionary history. Firstly, where the stratigraphic first occurrence dates of terminals (species,
genera and higher taxa) are known with reasonable accuracy, the extent of implied ghosts lineages
(relative to theoretical limits) offers an index of the 'goodness of fit' to the fossil record. Secondly, the
biogeographical or palaeobiogeographical distributions of many groups contain a residual
evolutionary signal, and this can also be tested for its fit to competing trees. While neither stratigraphy
nor biogeography necessarily offers foolproof discrimination in any particular case, their application
to a large statistical sample of cases - different major clades, different taxonomic levels, different ages
of radiations - allows for some level of generality. The project will also identify which clades under
which circumstances are most prone to morphological convergence, and therefore most suitably
analysed with molecular data.
The student will acquire an interdisciplinary and transferrable skill set: bioinformatics, phylogenetics,
statistics and palaeontology.