Wednesday, May 7, 2008

Nested hierarchies

Also during the long discussion yesterday, the topic of nested hierarchies came up. This caused me to look harder into the topic, to try and understand it. I found something interesting.

First, the main point. Evolutionists use "nested hierarchies" as evidence of common descent -- that is, life can be categories along certain lines, and there are no characteristics that violate those hierarchal categorization. Therefore, those hierarchies constitute the "family tree" of life.

When I first heard the argument, I was struck by how "label over reality" it was. That is, we categorize life according to certain characteristics, and then treat the structure of those categories as being "real." When if we picked different characteristics, we'd certainly have different hierarchies.

In fact, it seemed, there are all sorts of traits that violate the "rule of nested hierarchies."

So with this in mind, I started reading about the marsupials and monotremes, because they seemed to be a great place to start in looking into these hierarchies.

And what did I find? I found placental and marsupial variants of a bunch of different animals. All of the photos on this page are of MARSUPIALS, either alive or extinct. Yet we know instinctively, from looking at them, that they are eerily reminiscent of placental counterparts.

For example, there is the placental wolf, and the "tasmanian wolf," which was hunted to extinction in the 30s. The two species share all the major "dog" characteristics, except radically different reproductive systems, and, apparently, minor differences in the palate bone. So I kept looking, and I found the "marsupial sabertooth tiger," the "marsupial lion," the "marsupial anteater," the "marsupial flying squirrel," the "marsupial mole" the "marsupial badger (or tasmanian devil), the "marsupial mouse" ... and it didn't appear to be ending anytime soon.

All the sources I found treat all these animals as "convergent evolution" -- that is, the "tasmanian wolf" is not a "wolf" -- it is a "marsupial that happened to evolve a lot of the same characteristics as wolves, because it filled the same niche." Maybe I could buy this once. But how many times? Dogs, lions, sabertooth tigers, anteaters, moles, flying squirrels?

How many times did this "convergent evolution" happen?
Or maybe these "nested hierarchies" aren't as involable as argued. Maybe these organisms originated fully formed, in two distinct variants. Or maybe they originated in a single, primal form with the capacity for both forms of reproduction in the gene pool, which "fixed" due to genetic drift over time?

Whatever explanation you choose, these "nested hierarchies" start to look pretty silly when you have such a long list of organisms that have such similar placental counterparts, with only one significant difference: the reproductive system.

(I got all the photos straight off Google Images, and claim Fair Use, as this is a not-for-profit, educational purpose that uses documents only in small part).


Matt Ackerman said...

I don't want to clutter up your blog with comments, but you seem to be responding amicably enough, so I will leave this skeleton of an argument, and if you are interested we can continue. Clearly if we consider all traits, all traits cannot possibly fit in a nested hierarchy. Consider blue eyes. Some cats have blue eyes. some humans have blue eyes. So if we group all blue eyed creatures together we are grouping some humans with some cats. But if our objective is to group together organisms which in some intuitive sense are most similar than a grouping of Blue eyed vs non blue eyed creatures would be extremely unsatisfactory.

So, why don't we just make mathematically explicit what our intuitive sense might tell us? Instead of considering a single character, or even a small number of characters, why don't we just consider every dang characteristic we can think of? Then, there is by mathematical necessity some nested hierarchy which assigns the minimum number of "unnested" traits (like 'dog-like-ness' in the marsupial vs. placental example)

Since such a tree exist by mathematical necessity it's mere existence proves nothing.

The salient question is weather trees produced with a small subset of the total traits are identical. In other words, is the tree I get with traits A,B, and C similar or identical to the tree produced with traits D,E and F? It is theoretically extremely implausible that any single trait would accurately estimate the "true" tree (if a true tree exists), but as we increase the number of traits considered our estimates of the true tree should converge, IF AND ONLY IF the true tree exists.

In other words if I consider all programs, as I increase the number of traits considered some true groupings might emerge, (Perhaps all internet browsers share allot of similarities) but increasing numbers of traits estimate the tree differently.

In phylogenetic analysis this is referred to as a boot-strapping estimate, or some times measure of support. If your conclusions are highly dependent on what traits you chose to look at, then you cannot be sure that your proposed tree represents something found in nature. However, high bootstrapping values are indicative that there is one true hierarchy which exists.

ungtss said...

interesting point -- I think the key to performing such an analysis would be defining our "traits" correctly -- "doglikeness" contains a number of more specific traits, that could be grouped differently, yielding different numbers for comparison. Seems like the most mathematically rigourous way to do that would be to measure the similarity of the respective genomes -- to see whether the marsupial is more similar to the placental mouse or the koala. Maybe they'll get to that some day.

ungtss said...

marsupial --> marsupial mouse