More on the 'cladistics' of sequences

[Jan De Laet]

Pierre:

Farris pointed out in 1972 (Estimating phylogenetic trees from distance
matrices. American Naturalist 106: 645-668; see p. 657; the relevant
passage is also quoted in full in Nixon and Carpenter 1993 - On
outgroups. Cladistics 9: 413-426) that outgroups can be added to an
ingroup dataset to come to a hypothesis of ingroup root after the
globally most parsimonious unrooted trees have been obtained for the
dataset that includes the outgroups; and that failure to find an ingroup
root in that way (because there is no single branch that separates
ingroup terminals and outgroup terminals ) warrants the conclusion that
the data used do not support the hypothesis that the ingroup is
monophyletic. In this, inclusion of multiple putative outgroups would be
just an application of the general principle to use as much relevant
data as possible. When again was the manual of PAUP published?

Best

Jan De Laet
Freshwater Biology
Royal Belgian Institute of Natural Sciences
Brussels, Belgium

Pierre deleporte wrote:

>> Relationships in the study group are first determined in an unrooted
>> tree
>> without character polarization. The outgroup is used only to root the
>> unrooted
>> tree, and at that time the characters are polarized but the
>> relationships in the
>> study group remain as they had been in the unrooted tree.
>>
>> Since a clade has only one root, all outgroups should root in the same
>> location on the unrooted tree regardless of the outgroup chosen. It
>> is certainly
>> true the the more distant an outgroup is from the study group, the
>> more difficult
>> it is to establish homologies, but that is a question of application not
>> theory.
>
>
> Swofford dealt with this problem in the manual of PAUP, and Farris is
> said to have come to the same conclusion independently, as follows:
> given that one can always make a mistake in choosing "the" outgroup
> (i.e. choose a taxon that is in fact a member of the ingroup at
> stake), it's preferable to take into account several putative
> outgroups in the analysis, in order to try and minimize the risk.
> If they all fall into the same place on the optimal topology, there is
> no ambiguity in rooting. If not, then it cannot be said for sure which
> rooting is right or wrong, but at least it's certain that some mistake
> has been committed, and one should better enlarge the scope of the
> analysis, enlarge the putative ingroup, and consider another series of
> putative outgroups for analysing this enlarged putative ingroup.
>
> This notwithstanding possible improvement of the character data set,
> which of course can always happen to change the rooting of the outgroups.
>
> Pierre