Data from: Data set incongruence and correlated character evolution: an example of functional convergence in the hind-limbs of stifftail diving ducks

The unwitting inclusion of convergent characters in phylogenetic estimates
poses a serious problem for efforts to recover phylogeny. Convergence is
not inscrutable, however, particularly when one group of characters tracks
phylogeny and another set tracks adaptive history. In such cases,
convergent characters may be correlated with one or a few functional
anatomical units and readily identifiable using comparative methods.
Stifftail ducks (Oxyurinae) offer one such opportunity to study correlated
character evolution and function in the context of phylogenetic
reconstruction. Morphological analyses place stifftail ducks as part of a
large clade of diving ducks that includes the sea ducks (Mergini),
Hymenolaimus, Merganetta, and Tachyeres, and possibly the pochards
(Aythyini). Molecular analyses, on the other hand, place stifftails far
from other diving ducks and suggest, moreover, that stifftails are
polyphyletic. Mitochondrial cytochrome b gene sequences of eight stifftail
species traditionally supposed to form a clade were compared with each
other and with sequences from 50 other anseriform and galliform species.
Stifftail ducks are not the sister group of sea ducks, but lie outside the
typical ducks (Anatinae). Of the four traditional stifftail genera,
monophyly of Oxyura and its sister group relationship with Nomonyx are
strongly supported. Heteronetta probably is the sister group of that
clade, but support is weak. Biziura is not a true stifftail. Within
Oxyura, Old World species (O. australis, O. leucocephala, O. maccoa)
appear to form a clade, with New World species (O. jamaicensis, O.
vittata) branching basally. Incongruence between molecules and morphology
is interpreted to be the result of adaptive specialization and functional
convergence in the hind limbs of Biziura and true stifftails. When
morphological characters are divided into classes, only hind-limb
characters are significantly in conflict with the molecular tree. Null
models of synonomous and nonsynonomous substitution based on patterns of
codon-degeneracy and chemical dissimilarity, likewise, indicate that the
nucleotide and amino acid changes postulated by the molecular tree are
more plausible than those postulated by the morphological tree. These
findings teach general lessons about the utility of highly adaptive
characters (in particular those related to foraging ecology) and
underscore the problems that convergence can pose for attempts to recover
phylogeny. They also demonstrate how the concept of natural data
partitions and simple models of evolution (e.g., parsimony, likelihood,
neutrality) can be used to test the accuracy of independent phylogenetic
estimates and provide arguments in favor of one tree topology over
another.