Re: Strange thoughts on PN - was Re: BAD vs. BADD

[gerarus@westnet.com.au]

>>Who knows, 100 years from now, we may be able to quantify
>>morphological differences in some way, and then we can define
>>Hominidae as everything that does not differ by more than 20
>>Morphological Units from homo sapiens, and dinosaurs as anything
>that
>>does not differ more than 100 MU from Iguanodon, and, thus,
>somewhere
>>in theropoda, there would be a boundary line and everything above it
>>would not be a dinosaur anymore. Sure, that's also arbitrary
>>(boundaries would shift if we switched 100 to 105), but all
>groupings
>>are.  

    While there is a fair amount of intuitive appeal in such a
concept (and such a classificatory system is essentially what lies at
the heart of phenetics), it suffers from being currently unworkable
(and probably will always be so). In order to classify taxa in such a
way, we would have to know pretty much everything about every taxon.
Because a totally phenetic classification carries no discrimination
between characters used in its composition and theoretically no
predictive value for characters as yet undescribed, it is actually
more vulnerable to upset from new characters identified in component
taxa than an evolutionary classification. Also there is a problem in
separating higher taxa according to 'morphological units' (assuming
such things could be identified) because the positions of boundaries
in a higher classification _do_ need to mean something according to
an external reference. Imagine the following (entirely theoretical
and abstract) example of a morphological continuum, where the letters
represent taxa and the dashes represent morphological units:
A---B---C---D. Each taxon is separated by three morphological units.
Now, imagine that the definition of what constitutes a 'family' is
set as a separation of five morphological units. Under this
definition, A and B would fall in a family doesn't include D, while C
and D fall in a family that doesn't include A. However, B and C
should also fall in a single family, as they are separated by less
than five morphological units from each other. Therefore, any attempt
to divide these taxa into families (or lumping them into a single
family) requires a violation of the rule that a family must include
all taxa separated by less than five units, and exclude all those
taxa separated by more than five. And if this seems too abstract,
consider lining up _Pisanosaurus_ - _Coelophysis_ - _Sinosauropteryx_
- _Dromaeosaurus_ - _Archaeopteryx_ - _Confuciusornis_ - _Anhima_ and
playing 'pin the boundary between the classes'. The division would be
arbitrary, yes, but the point is that there would be no reason
someone else couldn't place their chosen division elsewhere. This is
_not_ the same as the arbitrary designation of a unit of length to
represent a metre, because the metre is not something that requires
an outside reference. It does not matter if two 'metres' (i.e. a
length measured as being a metre) overlap. It _does_ matter if two
non-equivalent but not hierarchically-related (i.e. one does not fit
within another) classificatory units overlap.
    Actually, any classificatory system will be inherently flawed, as
we are trying to represent a four-dimensional structure (the
evolutionary history of organisms) in a two-dimensional form (a
strictly hierarchical classification) [right now, I'm kind of wishing
I had gotten around to reading _Flatland_]. I would suggest that
there are four factors (not necessarily exclusive) that could be
considered in constructing a classification, which I would call the
four D's - Diversity (the existence of separate taxa), Descent
(phylogenetic relationships), Disparity (amount of difference in
characters between taxa) and Divergence (time since separation).
Phylogenetic nomenclature prioritises Descent out of these four,
while a Linnean classification arguably emphasises Disparity, and as
I've already pointed out, Disparity (which has no objective reference
point) is a more unreliable basis for a classification than Descent
(which does have such a point in the Last Universal Common Ancestor).
    There would be one possibility to overcome the problem with
Disparity - if we arbitrarily fixed a reference point (say, _Homo
sapiens_), and said that Disparity should be measured as distance
from that point. However, this would completely fail to give us a way
of recognising the tree structure of relationships.

    Cheers,

        Christopher Taylor (who tends not to stick unswervingly to a
strictly phylogenetic classificatory system and has been known to use
paraphyletic taxa, but is then unable to think of a proper defensive
argument for his own position)