Under & Verboom, Generic limits in the Rytidosperma complex 
607 
Within this group four distinct lineages can be recognised. These lineages are retrieved 
by the smaller analysis: only one is distinct in the larger analysis, in which substantial 
resolution is lost. This loss of resolution could be due to the inclusion of taxa with 
different character combinations, thereby breaking up the characters that define the 
subclades. It is striking, though, that the obvious characters which define the clades 
do not have exceptions. This indicates that convergences in other characters break 
up these clades. 
The taxonomic rank of the four clades, relative to the broader Rytidosperma, is 
problematic, and they either be recognised as subgenera of a broad Rytidosperma s.l., 
or as distinct genera. The arguments for recognising a broad Rytidosperma s.l. are as 
follows: 
a) This is taxonomically conservative, as this is the sense in which Zotov and Connor 
and Edgar used the genus, and the grasses in Australasia that were previously 
referred to as Danthonia will now be called Rytidosperma. 
b) This usage is consistent with common-name usage, where all Australian species 
of Rytidosperma s.l. are referred to as 'Wallaby grasses'. The most common Australian 
Rytidosperma s.l. species belong to one clade, so that even with a fragmented 
classification, 'Wallaby grasses' would still largely refer to Notodanthonia s.s. 
c) It may result in fewer name changes than a finer subdivision of the genus. However, 
due to the name changes proposed by Veldkamp (1980) not many new combinations 
would result from the fragmentation of Rytidosperma. 
d) The node on which the genus will be based was retrieved by all the analyses with 
different combinations of taxa, thus there is a good likelihood that this node is 
phylogenetically robust. 
Arguments that can be mustered for recognising the four clades as genera are as follows: 
a) Rytidosperma s.l. is a large genus, including 72 species. This is unwieldy; smaller 
genera are easier to understand and to work on. Subdivision would establish genera 
of 28, 35, 4 and 5 species. 
b) The inclusion of Karroochloa under Rytidosperma appears unlikely from cytogenetic 
grounds. We suspect that its inclusion within the Rytidosperma complex may be due 
to convergence. Recognising the narrower genera avoids implicating taxonomic 
changes in phylogenetic changes. 
c) The four clades are eco-geographically quite distinct. If taxonomic ranks should 
be established at nodes at which they are maximally predictive, then the recognition 
of the four as genera, rather than one variable genus, might be more informative. 
This would be consistent with the ecogeographical differences between Joycea, 
Plinthanthesis, and Notochloe. 
Morphological support is rather equivocal. Rytidosperma s.l. can be diagnosed by the 
usually tufted lemma indumentum: but there are exceptions, and the definition of 
tuftedness would have to be precise to avoid confusion with the African Merxmuellera. 
Two of the smaller clades can be diagnosed by a single, unambiguous character, the 
remaining two by character combinations. Thus for both rankings, most species can 
be keyed to the genus without difficulty, but there are some species which are 
difficult to place. 
On the balance of the evidence, we recognise four segregate genera, but are aware 
that there is an almost equally strong case for recognising a single, large genus 
Rytidosperma. 
