Fwd: Re: Agreement yet?

[pierre deleporte]

At 22:36 17/07/2002 -0400, John Grehan wrote :
>(...) comments in response to Tom DiBenedetto :
>>>>If so, then I wonder from where comes the assumption that closest
>>>>geographical proximity is evidence of closest relationship.
>>>
>>>Its not a matter that closest geographic proximity IS evidence of closest
>>>relationships. It provides a geographic criterion of a possible
>>>relationship according to the spatial criterion being applied. I don't
>>>know how to argue the case in any definitive terms. Is it true that in
>>>general for a vicariant series of related taxa the nearest neighbors have
>>>a closer biological relationship than those further away? Has anyone done
>>>a statistical test on this phenomenon?

This is in fact a possible way to charge tracks with some meaning. The kind
of assumptions I suggested an explanatory approach must implement in order
to make sense (models, explanatory laws...). This model is effectively
testable. Just note that for doing this, you must first rely on phylogeny,
then check for spatial distribution of already identified sister groups.
Your approach would then be:
1) assume the phylogeny of some groups is known,
2) test for the generality of the "phylogenetic/distributional law" in
these groups
3) generalize to groups of known distribution but unknown phylogeny -or
whose phylogeny you decide to ignore).

The paradox is that point 2 (the very basis of your model) could be
sustained all along by documenting new phylogenies and distributions,
dispensing with point 3.
In other words, a test of some predictions from point 3 would again be
point 2 : a phylogeny-base approach.
As far as I can reason on this proposition...

>>>Unjustifiable or not the method has resulted in predictions of
>>>relationship that were later corroborated by new biological information.

Here again, it could be tested how often does it "work", and how often does
it not "work"... given known phylogenies, once more.

An objection would remain, that is that nothing in spatial pattern of a
"track" seems to permit distinction between "when it will work" and "when
it will not work", because closer is closer, period, with no richer data to
think about. Unless you refine your notion of spatial "proximity", which
would inevitably involve "models" of some kind...

More on the Pacific-Atlantic example (below) in a next post.

Pierre

>For a particular cladogram of relationships for a taxon one can map out the
>track and identify what baseline may be appropriate to classify the spatial
>homology. For example, one might have a cladogram each for two groups that
>are identical in arrangement (and thus suggesting to some a common history)
>yet one group has a track spanning the Pacific Ocean basin, the other
>spanning the Atlantic.
>John Grehan