THERIZINOSAURS,LOGIC AND BRAINS

["Jonathan R. Wagner" <znc14@TTACS.TTU.EDU>]

At 08:05 PM 7/30/98 -0400, George Olshevsky wrote:
>My argument
>against the facile acceptance of reversals stems from first principles; a true
>reversal is a derived state of a derived state in which the second derived
>state is the same as the primitive state.
        Saying it is "the same" is dubious. It might be better to say that
the second derived state is of such morphology that it must be scored as the
primitive state in the data matrix. Not all reversals show identical
morphology, nor should they be expected to. Indeed, we should not expect
homologous character states to exhibit the same morphology in all instances
(e.g. "presence of ossified ilium" in mammals, lepidosaurs, _Megalosaurus_,
hadrosaurs, etc.).

>Given the large morphospace of all possible character states, what is the
>likelihood that the second derived state will find exactly the same
morphology >as the primitive state?
        Probably low. As alluded to above, no one stipulates that the
morphology be *exactly* the same. All that we can recognize in the
morphology of animals is a character and its states. These states are, in a
sense, a classification of the morphology of the animal. "Tetradactyly" is a
character state which encompasses a wide range of morphologies and implied
functions. We score everything with four toes as having this character state
because we cannot a priori say that one instance is or is not homologous
with the others. The morphology may be very similar, or very different; what
matters is that the morphology fits within the description of the character,
implying the potential of homology. The only way we can determine if
tetradactyly is homologous in several taxa is by testing that homology using
a phylogenetic hypothesis.

>If a feature is >truly< bistate, this chance might be as high as 50%, assuming
>that a change occurs at all. But at 50% even a short series of bistate
>characters has a low probability of becoming completely reversed, since the
>probabilities multiply.
        As I state above, no one is specifying an *exact* return to the
original state.
        Further, I don't think your little thought experiment applies.
Evolution is not about random occurances, it is about the differential
survival of random occurances. Sure, there may be a .5^5 or .5^6 chance of
returning to an ancestral state (and I'm not sure I buy that, but we'll
stick with your model). However, your statements assumes RANDOM acceptance
of a change in the genome of a population. While natural selection may use
random genetic preturbations as fuel for the engine of change, it does not
*act* randomly. Indeed, the entire point of natural selection is to favor
those changes which are advantageous. Thus, although there may only be a 2%
or 3% chance of the mutation occurring (probably much less), each
incremental step will be favored under the correct selection pressure.
        It doesn't matter how unlikley the event is, simply how beneficial
it is. If you take your short series of 50% chances, each one being of
benefit to the organism, each one is likley to be expressed in future
generations. That's natural selection in action.
        As has been argued many times before, these character state
transformations may not be have the original 50% (or whatever) chance of
coming about that we see in novelties. Since the ancestral state was
possessed by an ancestor (obviously), it is possible that it is in fact
easier to re-acquire the ancestral state than to evolve a new state from
whole cloth. Thus, on the whole, it may be easier to reverse than to
converge. However, this may vary on a case by case basis, and may be
difficult to quantify in any case. This is why it is probably better to give
equal weight to both types of homoplasy.

>Most characters exhibit a continuum of states, and the
>probability of a serial reversal of such characters is astronomically small.
        Except that a reversal doesn't have to go through the *entire*
continuum (e.g. Digit I of therizinosaurs doesn't entirely reach the ankle).
Further, if the continuum is closley correlated (as it should be to be coded
as one character), it may be that each "subcharacter" is so intimately
associated with the others that it is a relatively straightforward process
to reverse each one in turn (simple example: Digit I is 75% of Digit II,
then 50%, then 25%, then 10%).

>If your space has more than three dimensions, the probability that you will
>return to your starting point via a random walk is zero (this is from an
>article by Dan Asimov--Isaac's nephew and an old college acquaintance of
>mine--in _The Sciences_ a few years back). How many dimensions does a
>morphospace have?
        Your analogy to a random walk in multidimensional space does not
apply, no matter how intimately associated evolution has become with the
term random walk. For example: if you cannot ever make your way home again
in multidimensional morphospace using a random walk, can two random walks
which start at the same point in morphospace ever intersect? Are you about
to tell us that convergence cannot occur either?
        In evolution at the level we are discussing, there are areas of
morphospace which will not be touched due to selective pressure. Therefore,
your walk through morphospace is not entirely random, although perhaps
unpredictable. Also, you do leave a trail of breadcrumbs behind you, called
your genome. Does that happen in Asimov's models?
        There, that's the best I can do, because I don't understand this
sort of mathematics very well. What I do understand is that analogies to
chaotic behaviour which do not take into account the nature of organisms are
bound to run into trouble.

>The basisphenoid swelling in _Erlikosaurus_ is located at the
>back of the braincase below the basioccipital process.
        Actually, if I recall correctly, it was originally illustrated as
being on the basisphenoid immediately posterior to the parasphenoid, such
that if you simply redrew the "suture" between the parasphenoid and
basisphenoid behind the pneumatic enlargement, it would look like an
ornithomimid braincase (with perhaps a shorter basisphenoid). Not such a big
move...

>If living tissue were indeed this evolutionarily plastic, [...]
        I do find it interesting that you have not attempted to refute the
argument that features may move from bone to bone. I find the movement of a
feature from one bone to an adjacent bone on a skull with "obliterated"
sutures to be no significant threat to morphological analyses.

        Wagner
----------------------------------------------------------------------------
    Jonathan R. Wagner, Dept. of Geosciences, TTU, Lubbock, TX 79409-1053
                    "...To fight legends." - Kosh Naranek